+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.org wrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a
drop
in edits between August and September...
Is it possible that we had data issues before that were causing us to
see an
issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing
data
that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards <
arichards@wikimedia.org>
wrote:
I guess it's tough to draw a baseline considering we only have a couple months worth of data - but perhaps a good metric for our goal is
returning
to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org
wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that became active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org
wrote:
Well, looks like IndiaSummer95 got blocked today... but I've reached
out
to Nick1372 and KurtWags3 via their user talk pages.
On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org wrote:
I hope this is not too stalker-y but I think it is interesting.
I looked briefly at the top 5 editors for both August [A] and September [B], the most prolific editor "Indiasummer95", 520 days registered, [1] made 2908 edits in August but only 694 in September. The second most profilic editor "La Avatar Korra", registered for 88 days [2] in September made 954 edits in August but only 379 in September. Number 3, Nick1372, member for 306d days [3] had 762 edits in August but only 445 in September, #4 was 이민혁, registered for 70 days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 days [5] (520 dropping to 102)
Out of these Indiasummer95 was the most interesting so I looked at
the
edits per page for August [C]: Tell_Mama_UK 66 Abel_Xavier 60 Melilla 56 Vitória_S.C. 44 History_of_Roman_Catholicism_in_Portugal 40 Knattspyrnufélag_Reykjavíkur 40 Robot_Wars_(TV_series) 38 F.C._Paços_de_Ferreira 32 Symbols_of_Portugal 32 Politics_of_Toledo_(Spain) 32
The same in September [D]: Alfonso_I_of_Asturias 21 Indiasummer95 13 Vitālijs_Astafjevs 12 A.C._Perugia_Calcio 10 Morangos_com_Açúcar 9 Azad_Ali 9 Million_Muslim_March 8 Islam_in_the_Czech_Republic 7 Szilárd_Németh 7 Matteo_Ferrari 6
Inspecting closing on the dates of his/her edits to Tell_Mama_UK
(page
id 39683645) [E] 2013-08-11 4 2013-08-15 56 2013-08-16 6
However if I look at the history page I don't see 56 edits from him/her on that page (I count 36 total on the entire page):
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
So.. that makes me wonder...
- was something broken around 15th August in EventLogging - were we
reporting successes that weren't and thus the data is skewed (e.g. AbuseFilter related) ? Is there a logical explanation for this or is something indeed weird - can anyone suggest a reason? 2) Is this people getting used to the editor/wikitext - making lots
of
mistakes early on and then getting better at using it? 3) Is it indeed seasonal / people getting bored after initial excitement of editing?
[1]
https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti...
[2]
https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi...
[3]
https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be...
[4] https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro...
이민혁?mobileaction=beta
[5]
https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
[A] SELECT event_username, count(*) from MobileWebEditing_5644223 where event_action = 'success' and timestamp LIKE '201308%' group by event_username [B] SELECT event_username, count(*) from MobileWebEditing_5644223 where event_action = 'success' and timestamp LIKE '201309%' group by event_username [C] SELECT enwiki.page.page_title, count(*) from MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
= event_pageId where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group
by
event_pageId [D] SELECT enwiki.page.page_title, count(*) from MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
= event_pageId where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group
by
event_pageId [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = 39683645 group by DATE(timestamp)
On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org
wrote:
> I'll try to block out some time to look at creating those two
graphs
> tomorrow. Let me know if you want to help. > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org > wrote: >> >> Adding Dario to this thread. >> >> So the main question is why is the number of edits going down from >> august >> to september while the number of unique editors is going up? >> >> Few theories: >> >> 1) Problem with data - seems unlikely per Jon's analysis. Also,
both
>> numbers come from the same data source. >> >> 2) Change in the makeup of editors - not conclusive from the data >> but it >> seems that the number of edits from new editors and experienced >> editors are >> following the same general pattern. >> >> 3) Change in edits per session - no analysis done towards this
yet.
>> >> I'd like to take a look at: >> >> 1) a daily graph over the same two month period on unique daily >> editors >> >> 2) a daily graph of edits per session over the same two month >> period >> >> Few things to keep in mind: >> >> -It seems that there is not a seasonality difference in desktop >> edits, but >> as pointed out by Steven Walling it's not necessarily true that >> desktop >> edits and mobile edits will follow the same seasonality patterns. >> >> -Seasonality also wouldn't account for the skew between unique >> editors and >> raw edits. >> >> -This effect seems fairly centered on enwiki, however mobile
captcha
>> support was introduced and may affect numbers for other projects
in
>> some off >> setting fashion >> >> -The spike in number of edits occurs roughly between Aug 12 - 30, >> particularly 23-30. In that time section level editing was >> introduced on >> August 27. >> >> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson <jrobson@wikimedia.org
>> wrote: >>> >>> If you look at my graphs this theory doesn't seem to hold.... >>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: >>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>>> mpinchuk@wikimedia.org >>>> wrote: >>>> > One theory: most of our enwiki power users tried mobile
editing
>>>> > early >>>> > on and >>>> > made their customarily high number of edits per session in >>>> > August, but >>>> > they >>>> > weren't retained on mobile and didn't stick around through >>>> > September. >>>> > Meanwhile, new users continue to flood in, but they only make
an
>>>> > edit >>>> > or two >>>> > and leave. If this is the case, it makes a strong argument for >>>> > working >>>> > on >>>> > things like mobile GettingStarted/KeepGoing to bump up early >>>> > engagement for >>>> > newbies, and article histories/contrib views and other power >>>> > user >>>> > features >>>> > for oldbies :) >>>> >>>> This could be an interesting theory. Kenan, what do you think >>>> about >>>> reaching out to a handful of those editors that dropped and ask >>>> them >>>> why? >>>> >>>> --tomasz >> >> >> >> >> -- >> >> Kenan Wang >> Product Manager, Mobile >> Wikimedia Foundation > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.org wrote: Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a drop in edits between August and September...
Is it possible that we had data issues before that were causing us to see an issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing data that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards arichards@wikimedia.org wrote:
I guess it's tough to draw a baseline considering we only have a couple months worth of data - but perhaps a good metric for our goal is returning to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that became active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org wrote:
Well, looks like IndiaSummer95 got blocked today... but I've reached out to Nick1372 and KurtWags3 via their user talk pages.
On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org wrote:
I hope this is not too stalker-y but I think it is interesting.
I looked briefly at the top 5 editors for both August [A] and September [B], the most prolific editor "Indiasummer95", 520 days registered, [1] made 2908 edits in August but only 694 in September. The second most profilic editor "La Avatar Korra", registered for 88 days [2] in September made 954 edits in August but only 379 in September. Number 3, Nick1372, member for 306d days [3] had 762 edits in August but only 445 in September, #4 was 이민혁, registered for 70 days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 days [5] (520 dropping to 102)
Out of these Indiasummer95 was the most interesting so I looked at the edits per page for August [C]: Tell_Mama_UK 66 Abel_Xavier 60 Melilla 56 Vitória_S.C. 44 History_of_Roman_Catholicism_in_Portugal 40 Knattspyrnufélag_Reykjavíkur 40 Robot_Wars_(TV_series) 38 F.C._Paços_de_Ferreira 32 Symbols_of_Portugal 32 Politics_of_Toledo_(Spain) 32
The same in September [D]: Alfonso_I_of_Asturias 21 Indiasummer95 13 Vitālijs_Astafjevs 12 A.C._Perugia_Calcio 10 Morangos_com_Açúcar 9 Azad_Ali 9 Million_Muslim_March 8 Islam_in_the_Czech_Republic 7 Szilárd_Németh 7 Matteo_Ferrari 6
Inspecting closing on the dates of his/her edits to Tell_Mama_UK (page id 39683645) [E] 2013-08-11 4 2013-08-15 56 2013-08-16 6
However if I look at the history page I don't see 56 edits from him/her on that page (I count 36 total on the entire page):
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
So.. that makes me wonder...
- was something broken around 15th August in EventLogging - were we
reporting successes that weren't and thus the data is skewed (e.g. AbuseFilter related) ? Is there a logical explanation for this or is something indeed weird - can anyone suggest a reason? 2) Is this people getting used to the editor/wikitext - making lots of mistakes early on and then getting better at using it? 3) Is it indeed seasonal / people getting bored after initial excitement of editing?
[1] https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... [2] https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... [3] https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... [4] https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... [5] https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
[A] SELECT event_username, count(*) from MobileWebEditing_5644223 where event_action = 'success' and timestamp LIKE '201308%' group by event_username [B] SELECT event_username, count(*) from MobileWebEditing_5644223 where event_action = 'success' and timestamp LIKE '201309%' group by event_username [C] SELECT enwiki.page.page_title, count(*) from MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id = event_pageId where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group by event_pageId [D] SELECT enwiki.page.page_title, count(*) from MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id = event_pageId where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group by event_pageId [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 where wiki = 'enwiki' and event_action = 'success' and event_userName = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = 39683645 group by DATE(timestamp)
On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org wrote: > I'll try to block out some time to look at creating those two graphs > tomorrow. Let me know if you want to help. > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org > wrote: >> >> Adding Dario to this thread. >> >> So the main question is why is the number of edits going down from >> august >> to september while the number of unique editors is going up? >> >> Few theories: >> >> 1) Problem with data - seems unlikely per Jon's analysis. Also, both >> numbers come from the same data source. >> >> 2) Change in the makeup of editors - not conclusive from the data >> but it >> seems that the number of edits from new editors and experienced >> editors are >> following the same general pattern. >> >> 3) Change in edits per session - no analysis done towards this yet. >> >> I'd like to take a look at: >> >> 1) a daily graph over the same two month period on unique daily >> editors >> >> 2) a daily graph of edits per session over the same two month >> period >> >> Few things to keep in mind: >> >> -It seems that there is not a seasonality difference in desktop >> edits, but >> as pointed out by Steven Walling it's not necessarily true that >> desktop >> edits and mobile edits will follow the same seasonality patterns. >> >> -Seasonality also wouldn't account for the skew between unique >> editors and >> raw edits. >> >> -This effect seems fairly centered on enwiki, however mobile captcha >> support was introduced and may affect numbers for other projects in >> some off >> setting fashion >> >> -The spike in number of edits occurs roughly between Aug 12 - 30, >> particularly 23-30. In that time section level editing was >> introduced on >> August 27. >> >> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson jrobson@wikimedia.org >> wrote: >>> >>> If you look at my graphs this theory doesn't seem to hold.... >>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: >>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>>> mpinchuk@wikimedia.org >>>> wrote: >>>> > One theory: most of our enwiki power users tried mobile editing >>>> > early >>>> > on and >>>> > made their customarily high number of edits per session in >>>> > August, but >>>> > they >>>> > weren't retained on mobile and didn't stick around through >>>> > September. >>>> > Meanwhile, new users continue to flood in, but they only make an >>>> > edit >>>> > or two >>>> > and leave. If this is the case, it makes a strong argument for >>>> > working >>>> > on >>>> > things like mobile GettingStarted/KeepGoing to bump up early >>>> > engagement for >>>> > newbies, and article histories/contrib views and other power >>>> > user >>>> > features >>>> > for oldbies :) >>>> >>>> This could be an interesting theory. Kenan, what do you think >>>> about >>>> reaching out to a handful of those editors that dropped and ask >>>> them >>>> why? >>>> >>>> --tomasz >> >> >> >> >> -- >> >> Kenan Wang >> Product Manager, Mobile >> Wikimedia Foundation > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.org wrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a
drop
in edits between August and September...
Is it possible that we had data issues before that were causing us to
see an
issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing
data
that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards <
arichards@wikimedia.org>
wrote:
I guess it's tough to draw a baseline considering we only have a couple months worth of data - but perhaps a good metric for our goal is
returning
to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org
wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that became active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org
wrote:
Well, looks like IndiaSummer95 got blocked today... but I've reached
out
to Nick1372 and KurtWags3 via their user talk pages.
On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org wrote: > > I hope this is not too stalker-y but I think it is interesting. > > I looked briefly at the top 5 editors for both August [A] and > September [B], the most prolific editor "Indiasummer95", 520 days > registered, [1] made 2908 edits in August but only 694 in September. > The second most profilic editor "La Avatar Korra", registered for 88 > days [2] in September made 954 edits in August but only 379 in > September. Number 3, Nick1372, member for 306d days [3] had 762
edits
> in August but only 445 in September, #4 was 이민혁, registered for 70 > days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 > days [5] (520 dropping to 102) > > Out of these Indiasummer95 was the most interesting so I looked at
the
> edits per page for August [C]: > Tell_Mama_UK 66 > Abel_Xavier 60 > Melilla 56 > Vitória_S.C. 44 > History_of_Roman_Catholicism_in_Portugal 40 > Knattspyrnufélag_Reykjavíkur 40 > Robot_Wars_(TV_series) 38 > F.C._Paços_de_Ferreira 32 > Symbols_of_Portugal 32 > Politics_of_Toledo_(Spain) 32 > > The same in September [D]: > Alfonso_I_of_Asturias 21 > Indiasummer95 13 > Vitālijs_Astafjevs 12 > A.C._Perugia_Calcio 10 > Morangos_com_Açúcar 9 > Azad_Ali 9 > Million_Muslim_March 8 > Islam_in_the_Czech_Republic 7 > Szilárd_Németh 7 > Matteo_Ferrari 6 > > Inspecting closing on the dates of his/her edits to Tell_Mama_UK
(page
> id 39683645) [E] > 2013-08-11 4 > 2013-08-15 56 > 2013-08-16 6 > > However if I look at the history page I don't see 56 edits from > him/her on that page (I count 36 total on the entire page): > >
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
> > So.. that makes me wonder... > 1) was something broken around 15th August in EventLogging - were we > reporting successes that weren't and thus the data is skewed (e.g. > AbuseFilter related) ? Is there a logical explanation for this or is > something indeed weird - can anyone suggest a reason? > 2) Is this people getting used to the editor/wikitext - making lots
of
> mistakes early on and then getting better at using it? > 3) Is it indeed seasonal / people getting bored after initial > excitement of editing? > > [1] >
https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti...
> [2] >
https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi...
> [3] >
https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be...
> [4] > https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro...
이민혁?mobileaction=beta
> [5] >
https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
> > [A] SELECT event_username, count(*) from MobileWebEditing_5644223 > where event_action = 'success' and timestamp LIKE '201308%' group by > event_username > [B] SELECT event_username, count(*) from MobileWebEditing_5644223 > where event_action = 'success' and timestamp LIKE '201309%' group by > event_username > [C] SELECT enwiki.page.page_title, count(*) from > MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
> event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group
by
> event_pageId > [D] SELECT enwiki.page.page_title, count(*) from > MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
> event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group
by
> event_pageId > [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 > where wiki = 'enwiki' and event_action = 'success' and
event_userName
> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = > 39683645 group by DATE(timestamp) > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org
wrote:
> > I'll try to block out some time to look at creating those two
graphs
> > tomorrow. Let me know if you want to help. > > > > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org > > wrote: > >> > >> Adding Dario to this thread. > >> > >> So the main question is why is the number of edits going down
from
> >> august > >> to september while the number of unique editors is going up? > >> > >> Few theories: > >> > >> 1) Problem with data - seems unlikely per Jon's analysis. Also,
both
> >> numbers come from the same data source. > >> > >> 2) Change in the makeup of editors - not conclusive from the data > >> but it > >> seems that the number of edits from new editors and experienced > >> editors are > >> following the same general pattern. > >> > >> 3) Change in edits per session - no analysis done towards this
yet.
> >> > >> I'd like to take a look at: > >> > >> 1) a daily graph over the same two month period on unique daily > >> editors > >> > >> 2) a daily graph of edits per session over the same two month > >> period > >> > >> Few things to keep in mind: > >> > >> -It seems that there is not a seasonality difference in desktop > >> edits, but > >> as pointed out by Steven Walling it's not necessarily true that > >> desktop > >> edits and mobile edits will follow the same seasonality patterns. > >> > >> -Seasonality also wouldn't account for the skew between unique > >> editors and > >> raw edits. > >> > >> -This effect seems fairly centered on enwiki, however mobile
captcha
> >> support was introduced and may affect numbers for other projects
in
> >> some off > >> setting fashion > >> > >> -The spike in number of edits occurs roughly between Aug 12 - 30, > >> particularly 23-30. In that time section level editing was > >> introduced on > >> August 27. > >> > >> > >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson <
jrobson@wikimedia.org>
> >> wrote: > >>> > >>> If you look at my graphs this theory doesn't seem to hold.... > >>> > >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: > >>>> > >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk > >>>> mpinchuk@wikimedia.org > >>>> wrote: > >>>> > One theory: most of our enwiki power users tried mobile
editing
> >>>> > early > >>>> > on and > >>>> > made their customarily high number of edits per session in > >>>> > August, but > >>>> > they > >>>> > weren't retained on mobile and didn't stick around through > >>>> > September. > >>>> > Meanwhile, new users continue to flood in, but they only
make an
> >>>> > edit > >>>> > or two > >>>> > and leave. If this is the case, it makes a strong argument
for
> >>>> > working > >>>> > on > >>>> > things like mobile GettingStarted/KeepGoing to bump up early > >>>> > engagement for > >>>> > newbies, and article histories/contrib views and other power > >>>> > user > >>>> > features > >>>> > for oldbies :) > >>>> > >>>> This could be an interesting theory. Kenan, what do you think > >>>> about > >>>> reaching out to a handful of those editors that dropped and ask > >>>> them > >>>> why? > >>>> > >>>> --tomasz > >> > >> > >> > >> > >> -- > >> > >> Kenan Wang > >> Product Manager, Mobile > >> Wikimedia Foundation > > > > > > > > > > -- > > > > Kenan Wang > > Product Manager, Mobile > > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a
drop
in edits between August and September...
Is it possible that we had data issues before that were causing us to
see an
issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing
data
that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards <
arichards@wikimedia.org>
wrote:
I guess it's tough to draw a baseline considering we only have a
couple
months worth of data - but perhaps a good metric for our goal is
returning
to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org
wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that
became
active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org
wrote:
> > Well, looks like IndiaSummer95 got blocked today... but I've
reached out
> to Nick1372 and KurtWags3 via their user talk pages. > > > On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org > wrote: >> >> I hope this is not too stalker-y but I think it is interesting. >> >> I looked briefly at the top 5 editors for both August [A] and >> September [B], the most prolific editor "Indiasummer95", 520 days >> registered, [1] made 2908 edits in August but only 694 in
September.
>> The second most profilic editor "La Avatar Korra", registered for
88
>> days [2] in September made 954 edits in August but only 379 in >> September. Number 3, Nick1372, member for 306d days [3] had 762
edits
>> in August but only 445 in September, #4 was 이민혁, registered for 70 >> days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 >> days [5] (520 dropping to 102) >> >> Out of these Indiasummer95 was the most interesting so I looked at
the
>> edits per page for August [C]: >> Tell_Mama_UK 66 >> Abel_Xavier 60 >> Melilla 56 >> Vitória_S.C. 44 >> History_of_Roman_Catholicism_in_Portugal 40 >> Knattspyrnufélag_Reykjavíkur 40 >> Robot_Wars_(TV_series) 38 >> F.C._Paços_de_Ferreira 32 >> Symbols_of_Portugal 32 >> Politics_of_Toledo_(Spain) 32 >> >> The same in September [D]: >> Alfonso_I_of_Asturias 21 >> Indiasummer95 13 >> Vitālijs_Astafjevs 12 >> A.C._Perugia_Calcio 10 >> Morangos_com_Açúcar 9 >> Azad_Ali 9 >> Million_Muslim_March 8 >> Islam_in_the_Czech_Republic 7 >> Szilárd_Németh 7 >> Matteo_Ferrari 6 >> >> Inspecting closing on the dates of his/her edits to Tell_Mama_UK
(page
>> id 39683645) [E] >> 2013-08-11 4 >> 2013-08-15 56 >> 2013-08-16 6 >> >> However if I look at the history page I don't see 56 edits from >> him/her on that page (I count 36 total on the entire page): >> >>
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
>> >> So.. that makes me wonder... >> 1) was something broken around 15th August in EventLogging - were
we
>> reporting successes that weren't and thus the data is skewed (e.g. >> AbuseFilter related) ? Is there a logical explanation for this or
is
>> something indeed weird - can anyone suggest a reason? >> 2) Is this people getting used to the editor/wikitext - making
lots of
>> mistakes early on and then getting better at using it? >> 3) Is it indeed seasonal / people getting bored after initial >> excitement of editing? >> >> [1] >>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti...
>> [2] >>
https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi...
>> [3] >>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be...
>> [4] >> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro...
이민혁?mobileaction=beta
>> [5] >>
https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
>> >> [A] SELECT event_username, count(*) from MobileWebEditing_5644223 >> where event_action = 'success' and timestamp LIKE '201308%' group
by
>> event_username >> [B] SELECT event_username, count(*) from MobileWebEditing_5644223 >> where event_action = 'success' and timestamp LIKE '201309%' group
by
>> event_username >> [C] SELECT enwiki.page.page_title, count(*) from >> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%'
group by
>> event_pageId >> [D] SELECT enwiki.page.page_title, count(*) from >> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%'
group by
>> event_pageId >> [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 >> where wiki = 'enwiki' and event_action = 'success' and
event_userName
>> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = >> 39683645 group by DATE(timestamp) >> >> >> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org
wrote:
>> > I'll try to block out some time to look at creating those two
graphs
>> > tomorrow. Let me know if you want to help. >> > >> > >> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org >> > wrote: >> >> >> >> Adding Dario to this thread. >> >> >> >> So the main question is why is the number of edits going down
from
>> >> august >> >> to september while the number of unique editors is going up? >> >> >> >> Few theories: >> >> >> >> 1) Problem with data - seems unlikely per Jon's analysis. Also,
both
>> >> numbers come from the same data source. >> >> >> >> 2) Change in the makeup of editors - not conclusive from the
data
>> >> but it >> >> seems that the number of edits from new editors and experienced >> >> editors are >> >> following the same general pattern. >> >> >> >> 3) Change in edits per session - no analysis done towards this
yet.
>> >> >> >> I'd like to take a look at: >> >> >> >> 1) a daily graph over the same two month period on unique daily >> >> editors >> >> >> >> 2) a daily graph of edits per session over the same two month >> >> period >> >> >> >> Few things to keep in mind: >> >> >> >> -It seems that there is not a seasonality difference in desktop >> >> edits, but >> >> as pointed out by Steven Walling it's not necessarily true that >> >> desktop >> >> edits and mobile edits will follow the same seasonality
patterns.
>> >> >> >> -Seasonality also wouldn't account for the skew between unique >> >> editors and >> >> raw edits. >> >> >> >> -This effect seems fairly centered on enwiki, however mobile
captcha
>> >> support was introduced and may affect numbers for other
projects in
>> >> some off >> >> setting fashion >> >> >> >> -The spike in number of edits occurs roughly between Aug 12 -
30,
>> >> particularly 23-30. In that time section level editing was >> >> introduced on >> >> August 27. >> >> >> >> >> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson <
jrobson@wikimedia.org>
>> >> wrote: >> >>> >> >>> If you look at my graphs this theory doesn't seem to hold.... >> >>> >> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org
wrote:
>> >>>> >> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >> >>>> mpinchuk@wikimedia.org >> >>>> wrote: >> >>>> > One theory: most of our enwiki power users tried mobile
editing
>> >>>> > early >> >>>> > on and >> >>>> > made their customarily high number of edits per session in >> >>>> > August, but >> >>>> > they >> >>>> > weren't retained on mobile and didn't stick around through >> >>>> > September. >> >>>> > Meanwhile, new users continue to flood in, but they only
make an
>> >>>> > edit >> >>>> > or two >> >>>> > and leave. If this is the case, it makes a strong argument
for
>> >>>> > working >> >>>> > on >> >>>> > things like mobile GettingStarted/KeepGoing to bump up early >> >>>> > engagement for >> >>>> > newbies, and article histories/contrib views and other power >> >>>> > user >> >>>> > features >> >>>> > for oldbies :) >> >>>> >> >>>> This could be an interesting theory. Kenan, what do you think >> >>>> about >> >>>> reaching out to a handful of those editors that dropped and
ask
>> >>>> them >> >>>> why? >> >>>> >> >>>> --tomasz >> >> >> >> >> >> >> >> >> >> -- >> >> >> >> Kenan Wang >> >> Product Manager, Mobile >> >> Wikimedia Foundation >> > >> > >> > >> > >> > -- >> > >> > Kenan Wang >> > Product Manager, Mobile >> > Wikimedia Foundation > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
* Is this graph updating properly now, by running this query: https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ? * I understand the latest graph is different from old versions of it. When was this difference noticed? * What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a
drop
in edits between August and September...
Is it possible that we had data issues before that were causing us to
see an
issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing
data
that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards <
arichards@wikimedia.org>
wrote:
I guess it's tough to draw a baseline considering we only have a
couple
months worth of data - but perhaps a good metric for our goal is
returning
to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org
wrote:
> > Thanks to Juliusz we now have a graph of mobile editors that became > active editors. We can see a drop in the number of editors that
became
> active: > >
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
> > > On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org
wrote:
>> >> Well, looks like IndiaSummer95 got blocked today... but I've
reached out
>> to Nick1372 and KurtWags3 via their user talk pages. >> >> >> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org >> wrote: >>> >>> I hope this is not too stalker-y but I think it is interesting. >>> >>> I looked briefly at the top 5 editors for both August [A] and >>> September [B], the most prolific editor "Indiasummer95", 520 days >>> registered, [1] made 2908 edits in August but only 694 in
September.
>>> The second most profilic editor "La Avatar Korra", registered for
88
>>> days [2] in September made 954 edits in August but only 379 in >>> September. Number 3, Nick1372, member for 306d days [3] had 762
edits
>>> in August but only 445 in September, #4 was 이민혁, registered for 70 >>> days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 >>> days [5] (520 dropping to 102) >>> >>> Out of these Indiasummer95 was the most interesting so I looked
at the
>>> edits per page for August [C]: >>> Tell_Mama_UK 66 >>> Abel_Xavier 60 >>> Melilla 56 >>> Vitória_S.C. 44 >>> History_of_Roman_Catholicism_in_Portugal 40 >>> Knattspyrnufélag_Reykjavíkur 40 >>> Robot_Wars_(TV_series) 38 >>> F.C._Paços_de_Ferreira 32 >>> Symbols_of_Portugal 32 >>> Politics_of_Toledo_(Spain) 32 >>> >>> The same in September [D]: >>> Alfonso_I_of_Asturias 21 >>> Indiasummer95 13 >>> Vitālijs_Astafjevs 12 >>> A.C._Perugia_Calcio 10 >>> Morangos_com_Açúcar 9 >>> Azad_Ali 9 >>> Million_Muslim_March 8 >>> Islam_in_the_Czech_Republic 7 >>> Szilárd_Németh 7 >>> Matteo_Ferrari 6 >>> >>> Inspecting closing on the dates of his/her edits to Tell_Mama_UK
(page
>>> id 39683645) [E] >>> 2013-08-11 4 >>> 2013-08-15 56 >>> 2013-08-16 6 >>> >>> However if I look at the history page I don't see 56 edits from >>> him/her on that page (I count 36 total on the entire page): >>> >>>
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
>>> >>> So.. that makes me wonder... >>> 1) was something broken around 15th August in EventLogging - were
we
>>> reporting successes that weren't and thus the data is skewed (e.g. >>> AbuseFilter related) ? Is there a logical explanation for this or
is
>>> something indeed weird - can anyone suggest a reason? >>> 2) Is this people getting used to the editor/wikitext - making
lots of
>>> mistakes early on and then getting better at using it? >>> 3) Is it indeed seasonal / people getting bored after initial >>> excitement of editing? >>> >>> [1] >>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti...
>>> [2] >>>
https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi...
>>> [3] >>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be...
>>> [4] >>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro...
이민혁?mobileaction=beta
>>> [5] >>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
>>> >>> [A] SELECT event_username, count(*) from MobileWebEditing_5644223 >>> where event_action = 'success' and timestamp LIKE '201308%' group
by
>>> event_username >>> [B] SELECT event_username, count(*) from MobileWebEditing_5644223 >>> where event_action = 'success' and timestamp LIKE '201309%' group
by
>>> event_username >>> [C] SELECT enwiki.page.page_title, count(*) from >>> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>>> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
>>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%'
group by
>>> event_pageId >>> [D] SELECT enwiki.page.page_title, count(*) from >>> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>>> = event_pageId where wiki = 'enwiki' and event_action = 'success'
and
>>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%'
group by
>>> event_pageId >>> [E] SELECT DATE(timestamp), count(*) from
MobileWebEditing_5644223
>>> where wiki = 'enwiki' and event_action = 'success' and
event_userName
>>> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = >>> 39683645 group by DATE(timestamp) >>> >>> >>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org
wrote:
>>> > I'll try to block out some time to look at creating those two
graphs
>>> > tomorrow. Let me know if you want to help. >>> > >>> > >>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang <kwang@wikimedia.org
>>> > wrote: >>> >> >>> >> Adding Dario to this thread. >>> >> >>> >> So the main question is why is the number of edits going down
from
>>> >> august >>> >> to september while the number of unique editors is going up? >>> >> >>> >> Few theories: >>> >> >>> >> 1) Problem with data - seems unlikely per Jon's analysis.
Also, both
>>> >> numbers come from the same data source. >>> >> >>> >> 2) Change in the makeup of editors - not conclusive from the
data
>>> >> but it >>> >> seems that the number of edits from new editors and experienced >>> >> editors are >>> >> following the same general pattern. >>> >> >>> >> 3) Change in edits per session - no analysis done towards this
yet.
>>> >> >>> >> I'd like to take a look at: >>> >> >>> >> 1) a daily graph over the same two month period on unique daily >>> >> editors >>> >> >>> >> 2) a daily graph of edits per session over the same two month >>> >> period >>> >> >>> >> Few things to keep in mind: >>> >> >>> >> -It seems that there is not a seasonality difference in desktop >>> >> edits, but >>> >> as pointed out by Steven Walling it's not necessarily true that >>> >> desktop >>> >> edits and mobile edits will follow the same seasonality
patterns.
>>> >> >>> >> -Seasonality also wouldn't account for the skew between unique >>> >> editors and >>> >> raw edits. >>> >> >>> >> -This effect seems fairly centered on enwiki, however mobile
captcha
>>> >> support was introduced and may affect numbers for other
projects in
>>> >> some off >>> >> setting fashion >>> >> >>> >> -The spike in number of edits occurs roughly between Aug 12 -
30,
>>> >> particularly 23-30. In that time section level editing was >>> >> introduced on >>> >> August 27. >>> >> >>> >> >>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson <
jrobson@wikimedia.org>
>>> >> wrote: >>> >>> >>> >>> If you look at my graphs this theory doesn't seem to hold.... >>> >>> >>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org
wrote:
>>> >>>> >>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>> >>>> mpinchuk@wikimedia.org >>> >>>> wrote: >>> >>>> > One theory: most of our enwiki power users tried mobile
editing
>>> >>>> > early >>> >>>> > on and >>> >>>> > made their customarily high number of edits per session in >>> >>>> > August, but >>> >>>> > they >>> >>>> > weren't retained on mobile and didn't stick around through >>> >>>> > September. >>> >>>> > Meanwhile, new users continue to flood in, but they only
make an
>>> >>>> > edit >>> >>>> > or two >>> >>>> > and leave. If this is the case, it makes a strong argument
for
>>> >>>> > working >>> >>>> > on >>> >>>> > things like mobile GettingStarted/KeepGoing to bump up
early
>>> >>>> > engagement for >>> >>>> > newbies, and article histories/contrib views and other
power
>>> >>>> > user >>> >>>> > features >>> >>>> > for oldbies :) >>> >>>> >>> >>>> This could be an interesting theory. Kenan, what do you think >>> >>>> about >>> >>>> reaching out to a handful of those editors that dropped and
ask
>>> >>>> them >>> >>>> why? >>> >>>> >>> >>>> --tomasz >>> >> >>> >> >>> >> >>> >> >>> >> -- >>> >> >>> >> Kenan Wang >>> >> Product Manager, Mobile >>> >> Wikimedia Foundation >>> > >>> > >>> > >>> > >>> > -- >>> > >>> > Kenan Wang >>> > Product Manager, Mobile >>> > Wikimedia Foundation >> >> >> >> >> -- >> >> Kenan Wang >> Product Manager, Mobile >> Wikimedia Foundation > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu dandreescu@wikimedia.orgwrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects
a drop
in edits between August and September...
Is it possible that we had data issues before that were causing us
to see an
issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be
missing data
that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards <
arichards@wikimedia.org>
wrote: > > I guess it's tough to draw a baseline considering we only have a
couple
> months worth of data - but perhaps a good metric for our goal is
returning
> to or exceeding our peak? > > > On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org
wrote:
>> >> Thanks to Juliusz we now have a graph of mobile editors that became >> active editors. We can see a drop in the number of editors that
became
>> active: >> >>
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
>> >> >> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org
wrote:
>>> >>> Well, looks like IndiaSummer95 got blocked today... but I've
reached out
>>> to Nick1372 and KurtWags3 via their user talk pages. >>> >>> >>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson <jrobson@wikimedia.org
>>> wrote: >>>> >>>> I hope this is not too stalker-y but I think it is interesting. >>>> >>>> I looked briefly at the top 5 editors for both August [A] and >>>> September [B], the most prolific editor "Indiasummer95", 520 days >>>> registered, [1] made 2908 edits in August but only 694 in
September.
>>>> The second most profilic editor "La Avatar Korra", registered
for 88
>>>> days [2] in September made 954 edits in August but only 379 in >>>> September. Number 3, Nick1372, member for 306d days [3] had 762
edits
>>>> in August but only 445 in September, #4 was 이민혁, registered for
70
>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered for
84
>>>> days [5] (520 dropping to 102) >>>> >>>> Out of these Indiasummer95 was the most interesting so I looked
at the
>>>> edits per page for August [C]: >>>> Tell_Mama_UK 66 >>>> Abel_Xavier 60 >>>> Melilla 56 >>>> Vitória_S.C. 44 >>>> History_of_Roman_Catholicism_in_Portugal 40 >>>> Knattspyrnufélag_Reykjavíkur 40 >>>> Robot_Wars_(TV_series) 38 >>>> F.C._Paços_de_Ferreira 32 >>>> Symbols_of_Portugal 32 >>>> Politics_of_Toledo_(Spain) 32 >>>> >>>> The same in September [D]: >>>> Alfonso_I_of_Asturias 21 >>>> Indiasummer95 13 >>>> Vitālijs_Astafjevs 12 >>>> A.C._Perugia_Calcio 10 >>>> Morangos_com_Açúcar 9 >>>> Azad_Ali 9 >>>> Million_Muslim_March 8 >>>> Islam_in_the_Czech_Republic 7 >>>> Szilárd_Németh 7 >>>> Matteo_Ferrari 6 >>>> >>>> Inspecting closing on the dates of his/her edits to Tell_Mama_UK
(page
>>>> id 39683645) [E] >>>> 2013-08-11 4 >>>> 2013-08-15 56 >>>> 2013-08-16 6 >>>> >>>> However if I look at the history page I don't see 56 edits from >>>> him/her on that page (I count 36 total on the entire page): >>>> >>>>
https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history
>>>> >>>> So.. that makes me wonder... >>>> 1) was something broken around 15th August in EventLogging -
were we
>>>> reporting successes that weren't and thus the data is skewed
(e.g.
>>>> AbuseFilter related) ? Is there a logical explanation for this
or is
>>>> something indeed weird - can anyone suggest a reason? >>>> 2) Is this people getting used to the editor/wikitext - making
lots of
>>>> mistakes early on and then getting better at using it? >>>> 3) Is it indeed seasonal / people getting bored after initial >>>> excitement of editing? >>>> >>>> [1] >>>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti...
>>>> [2] >>>>
https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi...
>>>> [3] >>>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be...
>>>> [4] >>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro...
이민혁?mobileaction=beta
>>>> [5] >>>>
https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b...
>>>> >>>> [A] SELECT event_username, count(*) from
MobileWebEditing_5644223
>>>> where event_action = 'success' and timestamp LIKE '201308%'
group by
>>>> event_username >>>> [B] SELECT event_username, count(*) from
MobileWebEditing_5644223
>>>> where event_action = 'success' and timestamp LIKE '201309%'
group by
>>>> event_username >>>> [C] SELECT enwiki.page.page_title, count(*) from >>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>>>> = event_pageId where wiki = 'enwiki' and event_action =
'success' and
>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%'
group by
>>>> event_pageId >>>> [D] SELECT enwiki.page.page_title, count(*) from >>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on
enwiki.page.page_id
>>>> = event_pageId where wiki = 'enwiki' and event_action =
'success' and
>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%'
group by
>>>> event_pageId >>>> [E] SELECT DATE(timestamp), count(*) from
MobileWebEditing_5644223
>>>> where wiki = 'enwiki' and event_action = 'success' and
event_userName
>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = >>>> 39683645 group by DATE(timestamp) >>>> >>>> >>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org
wrote:
>>>> > I'll try to block out some time to look at creating those two
graphs
>>>> > tomorrow. Let me know if you want to help. >>>> > >>>> > >>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang <
kwang@wikimedia.org>
>>>> > wrote: >>>> >> >>>> >> Adding Dario to this thread. >>>> >> >>>> >> So the main question is why is the number of edits going down
from
>>>> >> august >>>> >> to september while the number of unique editors is going up? >>>> >> >>>> >> Few theories: >>>> >> >>>> >> 1) Problem with data - seems unlikely per Jon's analysis.
Also, both
>>>> >> numbers come from the same data source. >>>> >> >>>> >> 2) Change in the makeup of editors - not conclusive from the
data
>>>> >> but it >>>> >> seems that the number of edits from new editors and
experienced
>>>> >> editors are >>>> >> following the same general pattern. >>>> >> >>>> >> 3) Change in edits per session - no analysis done towards
this yet.
>>>> >> >>>> >> I'd like to take a look at: >>>> >> >>>> >> 1) a daily graph over the same two month period on unique
daily
>>>> >> editors >>>> >> >>>> >> 2) a daily graph of edits per session over the same two month >>>> >> period >>>> >> >>>> >> Few things to keep in mind: >>>> >> >>>> >> -It seems that there is not a seasonality difference in
desktop
>>>> >> edits, but >>>> >> as pointed out by Steven Walling it's not necessarily true
that
>>>> >> desktop >>>> >> edits and mobile edits will follow the same seasonality
patterns.
>>>> >> >>>> >> -Seasonality also wouldn't account for the skew between unique >>>> >> editors and >>>> >> raw edits. >>>> >> >>>> >> -This effect seems fairly centered on enwiki, however mobile
captcha
>>>> >> support was introduced and may affect numbers for other
projects in
>>>> >> some off >>>> >> setting fashion >>>> >> >>>> >> -The spike in number of edits occurs roughly between Aug 12 -
30,
>>>> >> particularly 23-30. In that time section level editing was >>>> >> introduced on >>>> >> August 27. >>>> >> >>>> >> >>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson <
jrobson@wikimedia.org>
>>>> >> wrote: >>>> >>> >>>> >>> If you look at my graphs this theory doesn't seem to hold.... >>>> >>> >>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org
wrote:
>>>> >>>> >>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>>> >>>> mpinchuk@wikimedia.org >>>> >>>> wrote: >>>> >>>> > One theory: most of our enwiki power users tried mobile
editing
>>>> >>>> > early >>>> >>>> > on and >>>> >>>> > made their customarily high number of edits per session in >>>> >>>> > August, but >>>> >>>> > they >>>> >>>> > weren't retained on mobile and didn't stick around through >>>> >>>> > September. >>>> >>>> > Meanwhile, new users continue to flood in, but they only
make an
>>>> >>>> > edit >>>> >>>> > or two >>>> >>>> > and leave. If this is the case, it makes a strong
argument for
>>>> >>>> > working >>>> >>>> > on >>>> >>>> > things like mobile GettingStarted/KeepGoing to bump up
early
>>>> >>>> > engagement for >>>> >>>> > newbies, and article histories/contrib views and other
power
>>>> >>>> > user >>>> >>>> > features >>>> >>>> > for oldbies :) >>>> >>>> >>>> >>>> This could be an interesting theory. Kenan, what do you
think
>>>> >>>> about >>>> >>>> reaching out to a handful of those editors that dropped and
ask
>>>> >>>> them >>>> >>>> why? >>>> >>>> >>>> >>>> --tomasz >>>> >> >>>> >> >>>> >> >>>> >> >>>> >> -- >>>> >> >>>> >> Kenan Wang >>>> >> Product Manager, Mobile >>>> >> Wikimedia Foundation >>>> > >>>> > >>>> > >>>> > >>>> > -- >>>> > >>>> > Kenan Wang >>>> > Product Manager, Mobile >>>> > Wikimedia Foundation >>> >>> >>> >>> >>> -- >>> >>> Kenan Wang >>> Product Manager, Mobile >>> Wikimedia Foundation >> >> >> >> >> -- >> >> Kenan Wang >> Product Manager, Mobile >> Wikimedia Foundation > > > > > -- > Arthur Richards > Software Engineer, Mobile > [[User:Awjrichards]] > IRC: awjr > +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu dandreescu@wikimedia.orgwrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote:
Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote: > Hey everyone. The monthly successful edits graph no longer reflects a drop > in edits between August and September... > > Is it possible that we had data issues before that were causing us to see an > issue that wasn't there? > > OR > > Is there some sort of data issue now that is causing us to be missing data > that we used to be able to see? > > > On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards < arichards@wikimedia.org> > wrote: >> >> I guess it's tough to draw a baseline considering we only have a couple >> months worth of data - but perhaps a good metric for our goal is returning >> to or exceeding our peak? >> >> >> On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org wrote: >>> >>> Thanks to Juliusz we now have a graph of mobile editors that became >>> active editors. We can see a drop in the number of editors that became >>> active: >>> >>> http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors >>> >>> >>> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org wrote: >>>> >>>> Well, looks like IndiaSummer95 got blocked today... but I've reached out >>>> to Nick1372 and KurtWags3 via their user talk pages. >>>> >>>> >>>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson < jrobson@wikimedia.org> >>>> wrote: >>>>> >>>>> I hope this is not too stalker-y but I think it is interesting. >>>>> >>>>> I looked briefly at the top 5 editors for both August [A] and >>>>> September [B], the most prolific editor "Indiasummer95", 520 days >>>>> registered, [1] made 2908 edits in August but only 694 in September. >>>>> The second most profilic editor "La Avatar Korra", registered for 88 >>>>> days [2] in September made 954 edits in August but only 379 in >>>>> September. Number 3, Nick1372, member for 306d days [3] had 762 edits >>>>> in August but only 445 in September, #4 was 이민혁, registered for 70 >>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 >>>>> days [5] (520 dropping to 102) >>>>> >>>>> Out of these Indiasummer95 was the most interesting so I looked at the >>>>> edits per page for August [C]: >>>>> Tell_Mama_UK 66 >>>>> Abel_Xavier 60 >>>>> Melilla 56 >>>>> Vitória_S.C. 44 >>>>> History_of_Roman_Catholicism_in_Portugal 40 >>>>> Knattspyrnufélag_Reykjavíkur 40 >>>>> Robot_Wars_(TV_series) 38 >>>>> F.C._Paços_de_Ferreira 32 >>>>> Symbols_of_Portugal 32 >>>>> Politics_of_Toledo_(Spain) 32 >>>>> >>>>> The same in September [D]: >>>>> Alfonso_I_of_Asturias 21 >>>>> Indiasummer95 13 >>>>> Vitālijs_Astafjevs 12 >>>>> A.C._Perugia_Calcio 10 >>>>> Morangos_com_Açúcar 9 >>>>> Azad_Ali 9 >>>>> Million_Muslim_March 8 >>>>> Islam_in_the_Czech_Republic 7 >>>>> Szilárd_Németh 7 >>>>> Matteo_Ferrari 6 >>>>> >>>>> Inspecting closing on the dates of his/her edits to Tell_Mama_UK (page >>>>> id 39683645) [E] >>>>> 2013-08-11 4 >>>>> 2013-08-15 56 >>>>> 2013-08-16 6 >>>>> >>>>> However if I look at the history page I don't see 56 edits from >>>>> him/her on that page (I count 36 total on the entire page): >>>>> >>>>> https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history >>>>> >>>>> So.. that makes me wonder... >>>>> 1) was something broken around 15th August in EventLogging - were we >>>>> reporting successes that weren't and thus the data is skewed (e.g. >>>>> AbuseFilter related) ? Is there a logical explanation for this or is >>>>> something indeed weird - can anyone suggest a reason? >>>>> 2) Is this people getting used to the editor/wikitext - making lots of >>>>> mistakes early on and then getting better at using it? >>>>> 3) Is it indeed seasonal / people getting bored after initial >>>>> excitement of editing? >>>>> >>>>> [1] >>>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... >>>>> [2] >>>>> https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... >>>>> [3] >>>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... >>>>> [4] >>>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... 이민혁?mobileaction=beta >>>>> [5] >>>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... >>>>> >>>>> [A] SELECT event_username, count(*) from MobileWebEditing_5644223 >>>>> where event_action = 'success' and timestamp LIKE '201308%' group by >>>>> event_username >>>>> [B] SELECT event_username, count(*) from MobileWebEditing_5644223 >>>>> where event_action = 'success' and timestamp LIKE '201309%' group by >>>>> event_username >>>>> [C] SELECT enwiki.page.page_title, count(*) from >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id >>>>> = event_pageId where wiki = 'enwiki' and event_action = 'success' and >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group by >>>>> event_pageId >>>>> [D] SELECT enwiki.page.page_title, count(*) from >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id >>>>> = event_pageId where wiki = 'enwiki' and event_action = 'success' and >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group by >>>>> event_pageId >>>>> [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 >>>>> where wiki = 'enwiki' and event_action = 'success' and event_userName >>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = >>>>> 39683645 group by DATE(timestamp) >>>>> >>>>> >>>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org wrote: >>>>> > I'll try to block out some time to look at creating those two graphs >>>>> > tomorrow. Let me know if you want to help. >>>>> > >>>>> > >>>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < kwang@wikimedia.org> >>>>> > wrote: >>>>> >> >>>>> >> Adding Dario to this thread. >>>>> >> >>>>> >> So the main question is why is the number of edits going down from >>>>> >> august >>>>> >> to september while the number of unique editors is going up? >>>>> >> >>>>> >> Few theories: >>>>> >> >>>>> >> 1) Problem with data - seems unlikely per Jon's analysis. Also, both >>>>> >> numbers come from the same data source. >>>>> >> >>>>> >> 2) Change in the makeup of editors - not conclusive from the data >>>>> >> but it >>>>> >> seems that the number of edits from new editors and experienced >>>>> >> editors are >>>>> >> following the same general pattern. >>>>> >> >>>>> >> 3) Change in edits per session - no analysis done towards this yet. >>>>> >> >>>>> >> I'd like to take a look at: >>>>> >> >>>>> >> 1) a daily graph over the same two month period on unique daily >>>>> >> editors >>>>> >> >>>>> >> 2) a daily graph of edits per session over the same two month >>>>> >> period >>>>> >> >>>>> >> Few things to keep in mind: >>>>> >> >>>>> >> -It seems that there is not a seasonality difference in desktop >>>>> >> edits, but >>>>> >> as pointed out by Steven Walling it's not necessarily true that >>>>> >> desktop >>>>> >> edits and mobile edits will follow the same seasonality patterns. >>>>> >> >>>>> >> -Seasonality also wouldn't account for the skew between unique >>>>> >> editors and >>>>> >> raw edits. >>>>> >> >>>>> >> -This effect seems fairly centered on enwiki, however mobile captcha >>>>> >> support was introduced and may affect numbers for other projects in >>>>> >> some off >>>>> >> setting fashion >>>>> >> >>>>> >> -The spike in number of edits occurs roughly between Aug 12
- 30,
>>>>> >> particularly 23-30. In that time section level editing was >>>>> >> introduced on >>>>> >> August 27. >>>>> >> >>>>> >> >>>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson < jrobson@wikimedia.org> >>>>> >> wrote: >>>>> >>> >>>>> >>> If you look at my graphs this theory doesn't seem to hold.... >>>>> >>> >>>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: >>>>> >>>> >>>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>>>> >>>> mpinchuk@wikimedia.org >>>>> >>>> wrote: >>>>> >>>> > One theory: most of our enwiki power users tried mobile editing >>>>> >>>> > early >>>>> >>>> > on and >>>>> >>>> > made their customarily high number of edits per session in >>>>> >>>> > August, but >>>>> >>>> > they >>>>> >>>> > weren't retained on mobile and didn't stick around through >>>>> >>>> > September. >>>>> >>>> > Meanwhile, new users continue to flood in, but they only make an >>>>> >>>> > edit >>>>> >>>> > or two >>>>> >>>> > and leave. If this is the case, it makes a strong argument for >>>>> >>>> > working >>>>> >>>> > on >>>>> >>>> > things like mobile GettingStarted/KeepGoing to bump up early >>>>> >>>> > engagement for >>>>> >>>> > newbies, and article histories/contrib views and other power >>>>> >>>> > user >>>>> >>>> > features >>>>> >>>> > for oldbies :) >>>>> >>>> >>>>> >>>> This could be an interesting theory. Kenan, what do you think >>>>> >>>> about >>>>> >>>> reaching out to a handful of those editors that dropped and ask >>>>> >>>> them >>>>> >>>> why? >>>>> >>>> >>>>> >>>> --tomasz >>>>> >> >>>>> >> >>>>> >> >>>>> >> >>>>> >> -- >>>>> >> >>>>> >> Kenan Wang >>>>> >> Product Manager, Mobile >>>>> >> Wikimedia Foundation >>>>> > >>>>> > >>>>> > >>>>> > >>>>> > -- >>>>> > >>>>> > Kenan Wang >>>>> > Product Manager, Mobile >>>>> > Wikimedia Foundation >>>> >>>> >>>> >>>> >>>> -- >>>> >>>> Kenan Wang >>>> Product Manager, Mobile >>>> Wikimedia Foundation >>> >>> >>> >>> >>> -- >>> >>> Kenan Wang >>> Product Manager, Mobile >>> Wikimedia Foundation >> >> >> >> >> -- >> Arthur Richards >> Software Engineer, Mobile >> [[User:Awjrichards]] >> IRC: awjr >> +1-415-839-6885 x6687 > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu dandreescu@wikimedia.orgwrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote:
> Several theories: > 1) I guess this now reflects reversions/deletions of edits? > 2) We purge data collected before a certain date > > I suspect the former? > > On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org > wrote: > > Hey everyone. The monthly successful edits graph no longer > reflects a drop > > in edits between August and September... > > > > Is it possible that we had data issues before that were causing us > to see an > > issue that wasn't there? > > > > OR > > > > Is there some sort of data issue now that is causing us to be > missing data > > that we used to be able to see? > > > > > > On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards < > arichards@wikimedia.org> > > wrote: > >> > >> I guess it's tough to draw a baseline considering we only have a > couple > >> months worth of data - but perhaps a good metric for our goal is > returning > >> to or exceeding our peak? > >> > >> > >> On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org > wrote: > >>> > >>> Thanks to Juliusz we now have a graph of mobile editors that > became > >>> active editors. We can see a drop in the number of editors that > became > >>> active: > >>> > >>> > http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors > >>> > >>> > >>> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org > wrote: > >>>> > >>>> Well, looks like IndiaSummer95 got blocked today... but I've > reached out > >>>> to Nick1372 and KurtWags3 via their user talk pages. > >>>> > >>>> > >>>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson < > jrobson@wikimedia.org> > >>>> wrote: > >>>>> > >>>>> I hope this is not too stalker-y but I think it is interesting. > >>>>> > >>>>> I looked briefly at the top 5 editors for both August [A] and > >>>>> September [B], the most prolific editor "Indiasummer95", 520 > days > >>>>> registered, [1] made 2908 edits in August but only 694 in > September. > >>>>> The second most profilic editor "La Avatar Korra", registered > for 88 > >>>>> days [2] in September made 954 edits in August but only 379 in > >>>>> September. Number 3, Nick1372, member for 306d days [3] had > 762 edits > >>>>> in August but only 445 in September, #4 was 이민혁, registered > for 70 > >>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered > for 84 > >>>>> days [5] (520 dropping to 102) > >>>>> > >>>>> Out of these Indiasummer95 was the most interesting so I > looked at the > >>>>> edits per page for August [C]: > >>>>> Tell_Mama_UK 66 > >>>>> Abel_Xavier 60 > >>>>> Melilla 56 > >>>>> Vitória_S.C. 44 > >>>>> History_of_Roman_Catholicism_in_Portugal 40 > >>>>> Knattspyrnufélag_Reykjavíkur 40 > >>>>> Robot_Wars_(TV_series) 38 > >>>>> F.C._Paços_de_Ferreira 32 > >>>>> Symbols_of_Portugal 32 > >>>>> Politics_of_Toledo_(Spain) 32 > >>>>> > >>>>> The same in September [D]: > >>>>> Alfonso_I_of_Asturias 21 > >>>>> Indiasummer95 13 > >>>>> Vitālijs_Astafjevs 12 > >>>>> A.C._Perugia_Calcio 10 > >>>>> Morangos_com_Açúcar 9 > >>>>> Azad_Ali 9 > >>>>> Million_Muslim_March 8 > >>>>> Islam_in_the_Czech_Republic 7 > >>>>> Szilárd_Németh 7 > >>>>> Matteo_Ferrari 6 > >>>>> > >>>>> Inspecting closing on the dates of his/her edits to > Tell_Mama_UK (page > >>>>> id 39683645) [E] > >>>>> 2013-08-11 4 > >>>>> 2013-08-15 56 > >>>>> 2013-08-16 6 > >>>>> > >>>>> However if I look at the history page I don't see 56 edits from > >>>>> him/her on that page (I count 36 total on the entire page): > >>>>> > >>>>> > https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history > >>>>> > >>>>> So.. that makes me wonder... > >>>>> 1) was something broken around 15th August in EventLogging - > were we > >>>>> reporting successes that weren't and thus the data is skewed > (e.g. > >>>>> AbuseFilter related) ? Is there a logical explanation for this > or is > >>>>> something indeed weird - can anyone suggest a reason? > >>>>> 2) Is this people getting used to the editor/wikitext - making > lots of > >>>>> mistakes early on and then getting better at using it? > >>>>> 3) Is it indeed seasonal / people getting bored after initial > >>>>> excitement of editing? > >>>>> > >>>>> [1] > >>>>> > https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... > >>>>> [2] > >>>>> > https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... > >>>>> [3] > >>>>> > https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... > >>>>> [4] > >>>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... > 이민혁?mobileaction=beta > >>>>> [5] > >>>>> > https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... > >>>>> > >>>>> [A] SELECT event_username, count(*) from > MobileWebEditing_5644223 > >>>>> where event_action = 'success' and timestamp LIKE '201308%' > group by > >>>>> event_username > >>>>> [B] SELECT event_username, count(*) from > MobileWebEditing_5644223 > >>>>> where event_action = 'success' and timestamp LIKE '201309%' > group by > >>>>> event_username > >>>>> [C] SELECT enwiki.page.page_title, count(*) from > >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on > enwiki.page.page_id > >>>>> = event_pageId where wiki = 'enwiki' and event_action = > 'success' and > >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%' > group by > >>>>> event_pageId > >>>>> [D] SELECT enwiki.page.page_title, count(*) from > >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on > enwiki.page.page_id > >>>>> = event_pageId where wiki = 'enwiki' and event_action = > 'success' and > >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%' > group by > >>>>> event_pageId > >>>>> [E] SELECT DATE(timestamp), count(*) from > MobileWebEditing_5644223 > >>>>> where wiki = 'enwiki' and event_action = 'success' and > event_userName > >>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and > event_pageId = > >>>>> 39683645 group by DATE(timestamp) > >>>>> > >>>>> > >>>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < > kwang@wikimedia.org> wrote: > >>>>> > I'll try to block out some time to look at creating those > two graphs > >>>>> > tomorrow. Let me know if you want to help. > >>>>> > > >>>>> > > >>>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < > kwang@wikimedia.org> > >>>>> > wrote: > >>>>> >> > >>>>> >> Adding Dario to this thread. > >>>>> >> > >>>>> >> So the main question is why is the number of edits going > down from > >>>>> >> august > >>>>> >> to september while the number of unique editors is going up? > >>>>> >> > >>>>> >> Few theories: > >>>>> >> > >>>>> >> 1) Problem with data - seems unlikely per Jon's analysis. > Also, both > >>>>> >> numbers come from the same data source. > >>>>> >> > >>>>> >> 2) Change in the makeup of editors - not conclusive from > the data > >>>>> >> but it > >>>>> >> seems that the number of edits from new editors and > experienced > >>>>> >> editors are > >>>>> >> following the same general pattern. > >>>>> >> > >>>>> >> 3) Change in edits per session - no analysis done towards > this yet. > >>>>> >> > >>>>> >> I'd like to take a look at: > >>>>> >> > >>>>> >> 1) a daily graph over the same two month period on unique > daily > >>>>> >> editors > >>>>> >> > >>>>> >> 2) a daily graph of edits per session over the same two > month > >>>>> >> period > >>>>> >> > >>>>> >> Few things to keep in mind: > >>>>> >> > >>>>> >> -It seems that there is not a seasonality difference in > desktop > >>>>> >> edits, but > >>>>> >> as pointed out by Steven Walling it's not necessarily true > that > >>>>> >> desktop > >>>>> >> edits and mobile edits will follow the same seasonality > patterns. > >>>>> >> > >>>>> >> -Seasonality also wouldn't account for the skew between > unique > >>>>> >> editors and > >>>>> >> raw edits. > >>>>> >> > >>>>> >> -This effect seems fairly centered on enwiki, however > mobile captcha > >>>>> >> support was introduced and may affect numbers for other > projects in > >>>>> >> some off > >>>>> >> setting fashion > >>>>> >> > >>>>> >> -The spike in number of edits occurs roughly between Aug 12 > - 30, > >>>>> >> particularly 23-30. In that time section level editing was > >>>>> >> introduced on > >>>>> >> August 27. > >>>>> >> > >>>>> >> > >>>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson < > jrobson@wikimedia.org> > >>>>> >> wrote: > >>>>> >>> > >>>>> >>> If you look at my graphs this theory doesn't seem to > hold.... > >>>>> >>> > >>>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org > wrote: > >>>>> >>>> > >>>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk > >>>>> >>>> mpinchuk@wikimedia.org > >>>>> >>>> wrote: > >>>>> >>>> > One theory: most of our enwiki power users tried mobile > editing > >>>>> >>>> > early > >>>>> >>>> > on and > >>>>> >>>> > made their customarily high number of edits per session > in > >>>>> >>>> > August, but > >>>>> >>>> > they > >>>>> >>>> > weren't retained on mobile and didn't stick around > through > >>>>> >>>> > September. > >>>>> >>>> > Meanwhile, new users continue to flood in, but they > only make an > >>>>> >>>> > edit > >>>>> >>>> > or two > >>>>> >>>> > and leave. If this is the case, it makes a strong > argument for > >>>>> >>>> > working > >>>>> >>>> > on > >>>>> >>>> > things like mobile GettingStarted/KeepGoing to bump up > early > >>>>> >>>> > engagement for > >>>>> >>>> > newbies, and article histories/contrib views and other > power > >>>>> >>>> > user > >>>>> >>>> > features > >>>>> >>>> > for oldbies :) > >>>>> >>>> > >>>>> >>>> This could be an interesting theory. Kenan, what do you > think > >>>>> >>>> about > >>>>> >>>> reaching out to a handful of those editors that dropped > and ask > >>>>> >>>> them > >>>>> >>>> why? > >>>>> >>>> > >>>>> >>>> --tomasz > >>>>> >> > >>>>> >> > >>>>> >> > >>>>> >> > >>>>> >> -- > >>>>> >> > >>>>> >> Kenan Wang > >>>>> >> Product Manager, Mobile > >>>>> >> Wikimedia Foundation > >>>>> > > >>>>> > > >>>>> > > >>>>> > > >>>>> > -- > >>>>> > > >>>>> > Kenan Wang > >>>>> > Product Manager, Mobile > >>>>> > Wikimedia Foundation > >>>> > >>>> > >>>> > >>>> > >>>> -- > >>>> > >>>> Kenan Wang > >>>> Product Manager, Mobile > >>>> Wikimedia Foundation > >>> > >>> > >>> > >>> > >>> -- > >>> > >>> Kenan Wang > >>> Product Manager, Mobile > >>> Wikimedia Foundation > >> > >> > >> > >> > >> -- > >> Arthur Richards > >> Software Engineer, Mobile > >> [[User:Awjrichards]] > >> IRC: awjr > >> +1-415-839-6885 x6687 > > > > > > > > > > -- > > > > Kenan Wang > > Product Manager, Mobile > > Wikimedia Foundation >
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
So I worked out what's happening but not sure exactly why. The graph is actually more accurate now.
It seems there are duplicate rows in the database table where every column has the same value except id. So when you do a UNION as we recently started doing for the editing graphs which omits the id field you end up with non-distinct values which a UNION will throw away.
Take a look at: select * from MobileWebEditing_5644223 where timestamp = '20130806192404' and event_username = 'Bender235'
This gives two edits that were exactly the same apart from the id. Not sure why this is the case (will explore more next week)
On Fri, Dec 13, 2013 at 6:10 PM, Kenan Wang kwang@wikimedia.org wrote:
Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu <dandreescu@wikimedia.org
wrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.orgwrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.orgwrote:
Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
> Kenan – can you give me a pointer to the graph you are referring to > and confirm what data source you are using? If you’re relying on > EventLogging data (MobileWebEditing) the log is not affected by reverts or > page deletions. > > On Dec 2, 2013, at 11:53 AM, Arthur Richards < > arichards@wikimedia.org> wrote: > > +analytics > > > On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote: > >> Several theories: >> 1) I guess this now reflects reversions/deletions of edits? >> 2) We purge data collected before a certain date >> >> I suspect the former? >> >> On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org >> wrote: >> > Hey everyone. The monthly successful edits graph no longer >> reflects a drop >> > in edits between August and September... >> > >> > Is it possible that we had data issues before that were causing >> us to see an >> > issue that wasn't there? >> > >> > OR >> > >> > Is there some sort of data issue now that is causing us to be >> missing data >> > that we used to be able to see? >> > >> > >> > On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards < >> arichards@wikimedia.org> >> > wrote: >> >> >> >> I guess it's tough to draw a baseline considering we only have a >> couple >> >> months worth of data - but perhaps a good metric for our goal is >> returning >> >> to or exceeding our peak? >> >> >> >> >> >> On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org >> wrote: >> >>> >> >>> Thanks to Juliusz we now have a graph of mobile editors that >> became >> >>> active editors. We can see a drop in the number of editors that >> became >> >>> active: >> >>> >> >>> >> http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors >> >>> >> >>> >> >>> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang < >> kwang@wikimedia.org> wrote: >> >>>> >> >>>> Well, looks like IndiaSummer95 got blocked today... but I've >> reached out >> >>>> to Nick1372 and KurtWags3 via their user talk pages. >> >>>> >> >>>> >> >>>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson < >> jrobson@wikimedia.org> >> >>>> wrote: >> >>>>> >> >>>>> I hope this is not too stalker-y but I think it is >> interesting. >> >>>>> >> >>>>> I looked briefly at the top 5 editors for both August [A] and >> >>>>> September [B], the most prolific editor "Indiasummer95", 520 >> days >> >>>>> registered, [1] made 2908 edits in August but only 694 in >> September. >> >>>>> The second most profilic editor "La Avatar Korra", registered >> for 88 >> >>>>> days [2] in September made 954 edits in August but only 379 in >> >>>>> September. Number 3, Nick1372, member for 306d days [3] had >> 762 edits >> >>>>> in August but only 445 in September, #4 was 이민혁, registered >> for 70 >> >>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered >> for 84 >> >>>>> days [5] (520 dropping to 102) >> >>>>> >> >>>>> Out of these Indiasummer95 was the most interesting so I >> looked at the >> >>>>> edits per page for August [C]: >> >>>>> Tell_Mama_UK 66 >> >>>>> Abel_Xavier 60 >> >>>>> Melilla 56 >> >>>>> Vitória_S.C. 44 >> >>>>> History_of_Roman_Catholicism_in_Portugal 40 >> >>>>> Knattspyrnufélag_Reykjavíkur 40 >> >>>>> Robot_Wars_(TV_series) 38 >> >>>>> F.C._Paços_de_Ferreira 32 >> >>>>> Symbols_of_Portugal 32 >> >>>>> Politics_of_Toledo_(Spain) 32 >> >>>>> >> >>>>> The same in September [D]: >> >>>>> Alfonso_I_of_Asturias 21 >> >>>>> Indiasummer95 13 >> >>>>> Vitālijs_Astafjevs 12 >> >>>>> A.C._Perugia_Calcio 10 >> >>>>> Morangos_com_Açúcar 9 >> >>>>> Azad_Ali 9 >> >>>>> Million_Muslim_March 8 >> >>>>> Islam_in_the_Czech_Republic 7 >> >>>>> Szilárd_Németh 7 >> >>>>> Matteo_Ferrari 6 >> >>>>> >> >>>>> Inspecting closing on the dates of his/her edits to >> Tell_Mama_UK (page >> >>>>> id 39683645) [E] >> >>>>> 2013-08-11 4 >> >>>>> 2013-08-15 56 >> >>>>> 2013-08-16 6 >> >>>>> >> >>>>> However if I look at the history page I don't see 56 edits >> from >> >>>>> him/her on that page (I count 36 total on the entire page): >> >>>>> >> >>>>> >> https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history >> >>>>> >> >>>>> So.. that makes me wonder... >> >>>>> 1) was something broken around 15th August in EventLogging - >> were we >> >>>>> reporting successes that weren't and thus the data is skewed >> (e.g. >> >>>>> AbuseFilter related) ? Is there a logical explanation for >> this or is >> >>>>> something indeed weird - can anyone suggest a reason? >> >>>>> 2) Is this people getting used to the editor/wikitext - >> making lots of >> >>>>> mistakes early on and then getting better at using it? >> >>>>> 3) Is it indeed seasonal / people getting bored after initial >> >>>>> excitement of editing? >> >>>>> >> >>>>> [1] >> >>>>> >> https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... >> >>>>> [2] >> >>>>> >> https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... >> >>>>> [3] >> >>>>> >> https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... >> >>>>> [4] >> >>>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... >> 이민혁?mobileaction=beta >> >>>>> [5] >> >>>>> >> https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... >> >>>>> >> >>>>> [A] SELECT event_username, count(*) from >> MobileWebEditing_5644223 >> >>>>> where event_action = 'success' and timestamp LIKE '201308%' >> group by >> >>>>> event_username >> >>>>> [B] SELECT event_username, count(*) from >> MobileWebEditing_5644223 >> >>>>> where event_action = 'success' and timestamp LIKE '201309%' >> group by >> >>>>> event_username >> >>>>> [C] SELECT enwiki.page.page_title, count(*) from >> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >> enwiki.page.page_id >> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >> 'success' and >> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201308%' >> group by >> >>>>> event_pageId >> >>>>> [D] SELECT enwiki.page.page_title, count(*) from >> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >> enwiki.page.page_id >> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >> 'success' and >> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE '201309%' >> group by >> >>>>> event_pageId >> >>>>> [E] SELECT DATE(timestamp), count(*) from >> MobileWebEditing_5644223 >> >>>>> where wiki = 'enwiki' and event_action = 'success' and >> event_userName >> >>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and >> event_pageId = >> >>>>> 39683645 group by DATE(timestamp) >> >>>>> >> >>>>> >> >>>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >> kwang@wikimedia.org> wrote: >> >>>>> > I'll try to block out some time to look at creating those >> two graphs >> >>>>> > tomorrow. Let me know if you want to help. >> >>>>> > >> >>>>> > >> >>>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >> kwang@wikimedia.org> >> >>>>> > wrote: >> >>>>> >> >> >>>>> >> Adding Dario to this thread. >> >>>>> >> >> >>>>> >> So the main question is why is the number of edits going >> down from >> >>>>> >> august >> >>>>> >> to september while the number of unique editors is going >> up? >> >>>>> >> >> >>>>> >> Few theories: >> >>>>> >> >> >>>>> >> 1) Problem with data - seems unlikely per Jon's analysis. >> Also, both >> >>>>> >> numbers come from the same data source. >> >>>>> >> >> >>>>> >> 2) Change in the makeup of editors - not conclusive from >> the data >> >>>>> >> but it >> >>>>> >> seems that the number of edits from new editors and >> experienced >> >>>>> >> editors are >> >>>>> >> following the same general pattern. >> >>>>> >> >> >>>>> >> 3) Change in edits per session - no analysis done towards >> this yet. >> >>>>> >> >> >>>>> >> I'd like to take a look at: >> >>>>> >> >> >>>>> >> 1) a daily graph over the same two month period on unique >> daily >> >>>>> >> editors >> >>>>> >> >> >>>>> >> 2) a daily graph of edits per session over the same two >> month >> >>>>> >> period >> >>>>> >> >> >>>>> >> Few things to keep in mind: >> >>>>> >> >> >>>>> >> -It seems that there is not a seasonality difference in >> desktop >> >>>>> >> edits, but >> >>>>> >> as pointed out by Steven Walling it's not necessarily true >> that >> >>>>> >> desktop >> >>>>> >> edits and mobile edits will follow the same seasonality >> patterns. >> >>>>> >> >> >>>>> >> -Seasonality also wouldn't account for the skew between >> unique >> >>>>> >> editors and >> >>>>> >> raw edits. >> >>>>> >> >> >>>>> >> -This effect seems fairly centered on enwiki, however >> mobile captcha >> >>>>> >> support was introduced and may affect numbers for other >> projects in >> >>>>> >> some off >> >>>>> >> setting fashion >> >>>>> >> >> >>>>> >> -The spike in number of edits occurs roughly between Aug >> 12 - 30, >> >>>>> >> particularly 23-30. In that time section level editing was >> >>>>> >> introduced on >> >>>>> >> August 27. >> >>>>> >> >> >>>>> >> >> >>>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson < >> jrobson@wikimedia.org> >> >>>>> >> wrote: >> >>>>> >>> >> >>>>> >>> If you look at my graphs this theory doesn't seem to >> hold.... >> >>>>> >>> >> >>>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org >> wrote: >> >>>>> >>>> >> >>>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >> >>>>> >>>> mpinchuk@wikimedia.org >> >>>>> >>>> wrote: >> >>>>> >>>> > One theory: most of our enwiki power users tried >> mobile editing >> >>>>> >>>> > early >> >>>>> >>>> > on and >> >>>>> >>>> > made their customarily high number of edits per >> session in >> >>>>> >>>> > August, but >> >>>>> >>>> > they >> >>>>> >>>> > weren't retained on mobile and didn't stick around >> through >> >>>>> >>>> > September. >> >>>>> >>>> > Meanwhile, new users continue to flood in, but they >> only make an >> >>>>> >>>> > edit >> >>>>> >>>> > or two >> >>>>> >>>> > and leave. If this is the case, it makes a strong >> argument for >> >>>>> >>>> > working >> >>>>> >>>> > on >> >>>>> >>>> > things like mobile GettingStarted/KeepGoing to bump up >> early >> >>>>> >>>> > engagement for >> >>>>> >>>> > newbies, and article histories/contrib views and other >> power >> >>>>> >>>> > user >> >>>>> >>>> > features >> >>>>> >>>> > for oldbies :) >> >>>>> >>>> >> >>>>> >>>> This could be an interesting theory. Kenan, what do you >> think >> >>>>> >>>> about >> >>>>> >>>> reaching out to a handful of those editors that dropped >> and ask >> >>>>> >>>> them >> >>>>> >>>> why? >> >>>>> >>>> >> >>>>> >>>> --tomasz >> >>>>> >> >> >>>>> >> >> >>>>> >> >> >>>>> >> >> >>>>> >> -- >> >>>>> >> >> >>>>> >> Kenan Wang >> >>>>> >> Product Manager, Mobile >> >>>>> >> Wikimedia Foundation >> >>>>> > >> >>>>> > >> >>>>> > >> >>>>> > >> >>>>> > -- >> >>>>> > >> >>>>> > Kenan Wang >> >>>>> > Product Manager, Mobile >> >>>>> > Wikimedia Foundation >> >>>> >> >>>> >> >>>> >> >>>> >> >>>> -- >> >>>> >> >>>> Kenan Wang >> >>>> Product Manager, Mobile >> >>>> Wikimedia Foundation >> >>> >> >>> >> >>> >> >>> >> >>> -- >> >>> >> >>> Kenan Wang >> >>> Product Manager, Mobile >> >>> Wikimedia Foundation >> >> >> >> >> >> >> >> >> >> -- >> >> Arthur Richards >> >> Software Engineer, Mobile >> >> [[User:Awjrichards]] >> >> IRC: awjr >> >> +1-415-839-6885 x6687 >> > >> > >> > >> > >> > -- >> > >> > Kenan Wang >> > Product Manager, Mobile >> > Wikimedia Foundation >> > > > > -- > Arthur Richards > Software Engineer, Mobile > [[User:Awjrichards]] > IRC: awjr > +1-415-839-6885 x6687 > > >
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
So Jon, Ori, and I agreed it'd be interesting to run the following queries:
duplicates in MobileWebEditing_5644223 including timestamp: about 315k, about 37k are the "success" event duplicates in MobileWebEditing_5644223 excluding timestamp: about 323k, about 39k are "success" duplicates in MobileWebEditing_6077315 including timestamp: 327, 4 "success" duplicates in MobileWebEditing_6077315 excluding timestamp: about 83k, about 3k are "success"
To get these I just grouped by the fields you guys are selecting in the query backing that graph, and got everything having count > 1. Of interest is that some things have count == 6. All the files are available in my home directory on stat1, I made them world readable. The queries were all like this:
select L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount, count(*) as duplicity from log.MobileWebEditing_6077315 L group by L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount having count(*) > 1 ;
Hm, I looked at one of these records and it seems not all the deduplications that happen because of the UNION are good. Some have different event_pageId values. So, when cleaning those tables, be careful to include that field and any others that matter.
On Fri, Dec 13, 2013 at 9:41 PM, Jon Robson jrobson@wikimedia.org wrote:
So I worked out what's happening but not sure exactly why. The graph is actually more accurate now.
It seems there are duplicate rows in the database table where every column has the same value except id. So when you do a UNION as we recently started doing for the editing graphs which omits the id field you end up with non-distinct values which a UNION will throw away.
Take a look at: select * from MobileWebEditing_5644223 where timestamp = '20130806192404' and event_username = 'Bender235'
This gives two edits that were exactly the same apart from the id. Not sure why this is the case (will explore more next week)
On Fri, Dec 13, 2013 at 6:10 PM, Kenan Wang kwang@wikimedia.org wrote:
Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu < dandreescu@wikimedia.org> wrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.orgwrote:
Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.orgwrote:
> Looking at this one: > http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful > > > On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < > dtaraborelli@wikimedia.org> wrote: > >> Kenan – can you give me a pointer to the graph you are referring to >> and confirm what data source you are using? If you’re relying on >> EventLogging data (MobileWebEditing) the log is not affected by reverts or >> page deletions. >> >> On Dec 2, 2013, at 11:53 AM, Arthur Richards < >> arichards@wikimedia.org> wrote: >> >> +analytics >> >> >> On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.orgwrote: >> >>> Several theories: >>> 1) I guess this now reflects reversions/deletions of edits? >>> 2) We purge data collected before a certain date >>> >>> I suspect the former? >>> >>> On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org >>> wrote: >>> > Hey everyone. The monthly successful edits graph no longer >>> reflects a drop >>> > in edits between August and September... >>> > >>> > Is it possible that we had data issues before that were causing >>> us to see an >>> > issue that wasn't there? >>> > >>> > OR >>> > >>> > Is there some sort of data issue now that is causing us to be >>> missing data >>> > that we used to be able to see? >>> > >>> > >>> > On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards < >>> arichards@wikimedia.org> >>> > wrote: >>> >> >>> >> I guess it's tough to draw a baseline considering we only have >>> a couple >>> >> months worth of data - but perhaps a good metric for our goal >>> is returning >>> >> to or exceeding our peak? >>> >> >>> >> >>> >> On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang < >>> kwang@wikimedia.org> wrote: >>> >>> >>> >>> Thanks to Juliusz we now have a graph of mobile editors that >>> became >>> >>> active editors. We can see a drop in the number of editors >>> that became >>> >>> active: >>> >>> >>> >>> >>> http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors >>> >>> >>> >>> >>> >>> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang < >>> kwang@wikimedia.org> wrote: >>> >>>> >>> >>>> Well, looks like IndiaSummer95 got blocked today... but I've >>> reached out >>> >>>> to Nick1372 and KurtWags3 via their user talk pages. >>> >>>> >>> >>>> >>> >>>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson < >>> jrobson@wikimedia.org> >>> >>>> wrote: >>> >>>>> >>> >>>>> I hope this is not too stalker-y but I think it is >>> interesting. >>> >>>>> >>> >>>>> I looked briefly at the top 5 editors for both August [A] and >>> >>>>> September [B], the most prolific editor "Indiasummer95", 520 >>> days >>> >>>>> registered, [1] made 2908 edits in August but only 694 in >>> September. >>> >>>>> The second most profilic editor "La Avatar Korra", >>> registered for 88 >>> >>>>> days [2] in September made 954 edits in August but only 379 >>> in >>> >>>>> September. Number 3, Nick1372, member for 306d days [3] had >>> 762 edits >>> >>>>> in August but only 445 in September, #4 was 이민혁, registered >>> for 70 >>> >>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered >>> for 84 >>> >>>>> days [5] (520 dropping to 102) >>> >>>>> >>> >>>>> Out of these Indiasummer95 was the most interesting so I >>> looked at the >>> >>>>> edits per page for August [C]: >>> >>>>> Tell_Mama_UK 66 >>> >>>>> Abel_Xavier 60 >>> >>>>> Melilla 56 >>> >>>>> Vitória_S.C. 44 >>> >>>>> History_of_Roman_Catholicism_in_Portugal 40 >>> >>>>> Knattspyrnufélag_Reykjavíkur 40 >>> >>>>> Robot_Wars_(TV_series) 38 >>> >>>>> F.C._Paços_de_Ferreira 32 >>> >>>>> Symbols_of_Portugal 32 >>> >>>>> Politics_of_Toledo_(Spain) 32 >>> >>>>> >>> >>>>> The same in September [D]: >>> >>>>> Alfonso_I_of_Asturias 21 >>> >>>>> Indiasummer95 13 >>> >>>>> Vitālijs_Astafjevs 12 >>> >>>>> A.C._Perugia_Calcio 10 >>> >>>>> Morangos_com_Açúcar 9 >>> >>>>> Azad_Ali 9 >>> >>>>> Million_Muslim_March 8 >>> >>>>> Islam_in_the_Czech_Republic 7 >>> >>>>> Szilárd_Németh 7 >>> >>>>> Matteo_Ferrari 6 >>> >>>>> >>> >>>>> Inspecting closing on the dates of his/her edits to >>> Tell_Mama_UK (page >>> >>>>> id 39683645) [E] >>> >>>>> 2013-08-11 4 >>> >>>>> 2013-08-15 56 >>> >>>>> 2013-08-16 6 >>> >>>>> >>> >>>>> However if I look at the history page I don't see 56 edits >>> from >>> >>>>> him/her on that page (I count 36 total on the entire page): >>> >>>>> >>> >>>>> >>> https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history >>> >>>>> >>> >>>>> So.. that makes me wonder... >>> >>>>> 1) was something broken around 15th August in EventLogging - >>> were we >>> >>>>> reporting successes that weren't and thus the data is skewed >>> (e.g. >>> >>>>> AbuseFilter related) ? Is there a logical explanation for >>> this or is >>> >>>>> something indeed weird - can anyone suggest a reason? >>> >>>>> 2) Is this people getting used to the editor/wikitext - >>> making lots of >>> >>>>> mistakes early on and then getting better at using it? >>> >>>>> 3) Is it indeed seasonal / people getting bored after initial >>> >>>>> excitement of editing? >>> >>>>> >>> >>>>> [1] >>> >>>>> >>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... >>> >>>>> [2] >>> >>>>> >>> https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... >>> >>>>> [3] >>> >>>>> >>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... >>> >>>>> [4] >>> >>>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... >>> 이민혁?mobileaction=beta >>> >>>>> [5] >>> >>>>> >>> https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... >>> >>>>> >>> >>>>> [A] SELECT event_username, count(*) from >>> MobileWebEditing_5644223 >>> >>>>> where event_action = 'success' and timestamp LIKE '201308%' >>> group by >>> >>>>> event_username >>> >>>>> [B] SELECT event_username, count(*) from >>> MobileWebEditing_5644223 >>> >>>>> where event_action = 'success' and timestamp LIKE '201309%' >>> group by >>> >>>>> event_username >>> >>>>> [C] SELECT enwiki.page.page_title, count(*) from >>> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >>> enwiki.page.page_id >>> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >>> 'success' and >>> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE >>> '201308%' group by >>> >>>>> event_pageId >>> >>>>> [D] SELECT enwiki.page.page_title, count(*) from >>> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >>> enwiki.page.page_id >>> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >>> 'success' and >>> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE >>> '201309%' group by >>> >>>>> event_pageId >>> >>>>> [E] SELECT DATE(timestamp), count(*) from >>> MobileWebEditing_5644223 >>> >>>>> where wiki = 'enwiki' and event_action = 'success' and >>> event_userName >>> >>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and >>> event_pageId = >>> >>>>> 39683645 group by DATE(timestamp) >>> >>>>> >>> >>>>> >>> >>>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >>> kwang@wikimedia.org> wrote: >>> >>>>> > I'll try to block out some time to look at creating those >>> two graphs >>> >>>>> > tomorrow. Let me know if you want to help. >>> >>>>> > >>> >>>>> > >>> >>>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >>> kwang@wikimedia.org> >>> >>>>> > wrote: >>> >>>>> >> >>> >>>>> >> Adding Dario to this thread. >>> >>>>> >> >>> >>>>> >> So the main question is why is the number of edits going >>> down from >>> >>>>> >> august >>> >>>>> >> to september while the number of unique editors is going >>> up? >>> >>>>> >> >>> >>>>> >> Few theories: >>> >>>>> >> >>> >>>>> >> 1) Problem with data - seems unlikely per Jon's analysis. >>> Also, both >>> >>>>> >> numbers come from the same data source. >>> >>>>> >> >>> >>>>> >> 2) Change in the makeup of editors - not conclusive from >>> the data >>> >>>>> >> but it >>> >>>>> >> seems that the number of edits from new editors and >>> experienced >>> >>>>> >> editors are >>> >>>>> >> following the same general pattern. >>> >>>>> >> >>> >>>>> >> 3) Change in edits per session - no analysis done towards >>> this yet. >>> >>>>> >> >>> >>>>> >> I'd like to take a look at: >>> >>>>> >> >>> >>>>> >> 1) a daily graph over the same two month period on unique >>> daily >>> >>>>> >> editors >>> >>>>> >> >>> >>>>> >> 2) a daily graph of edits per session over the same two >>> month >>> >>>>> >> period >>> >>>>> >> >>> >>>>> >> Few things to keep in mind: >>> >>>>> >> >>> >>>>> >> -It seems that there is not a seasonality difference in >>> desktop >>> >>>>> >> edits, but >>> >>>>> >> as pointed out by Steven Walling it's not necessarily >>> true that >>> >>>>> >> desktop >>> >>>>> >> edits and mobile edits will follow the same seasonality >>> patterns. >>> >>>>> >> >>> >>>>> >> -Seasonality also wouldn't account for the skew between >>> unique >>> >>>>> >> editors and >>> >>>>> >> raw edits. >>> >>>>> >> >>> >>>>> >> -This effect seems fairly centered on enwiki, however >>> mobile captcha >>> >>>>> >> support was introduced and may affect numbers for other >>> projects in >>> >>>>> >> some off >>> >>>>> >> setting fashion >>> >>>>> >> >>> >>>>> >> -The spike in number of edits occurs roughly between Aug >>> 12 - 30, >>> >>>>> >> particularly 23-30. In that time section level editing was >>> >>>>> >> introduced on >>> >>>>> >> August 27. >>> >>>>> >> >>> >>>>> >> >>> >>>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson < >>> jrobson@wikimedia.org> >>> >>>>> >> wrote: >>> >>>>> >>> >>> >>>>> >>> If you look at my graphs this theory doesn't seem to >>> hold.... >>> >>>>> >>> >>> >>>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org >>> wrote: >>> >>>>> >>>> >>> >>>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>> >>>>> >>>> mpinchuk@wikimedia.org >>> >>>>> >>>> wrote: >>> >>>>> >>>> > One theory: most of our enwiki power users tried >>> mobile editing >>> >>>>> >>>> > early >>> >>>>> >>>> > on and >>> >>>>> >>>> > made their customarily high number of edits per >>> session in >>> >>>>> >>>> > August, but >>> >>>>> >>>> > they >>> >>>>> >>>> > weren't retained on mobile and didn't stick around >>> through >>> >>>>> >>>> > September. >>> >>>>> >>>> > Meanwhile, new users continue to flood in, but they >>> only make an >>> >>>>> >>>> > edit >>> >>>>> >>>> > or two >>> >>>>> >>>> > and leave. If this is the case, it makes a strong >>> argument for >>> >>>>> >>>> > working >>> >>>>> >>>> > on >>> >>>>> >>>> > things like mobile GettingStarted/KeepGoing to bump >>> up early >>> >>>>> >>>> > engagement for >>> >>>>> >>>> > newbies, and article histories/contrib views and >>> other power >>> >>>>> >>>> > user >>> >>>>> >>>> > features >>> >>>>> >>>> > for oldbies :) >>> >>>>> >>>> >>> >>>>> >>>> This could be an interesting theory. Kenan, what do you >>> think >>> >>>>> >>>> about >>> >>>>> >>>> reaching out to a handful of those editors that dropped >>> and ask >>> >>>>> >>>> them >>> >>>>> >>>> why? >>> >>>>> >>>> >>> >>>>> >>>> --tomasz >>> >>>>> >> >>> >>>>> >> >>> >>>>> >> >>> >>>>> >> >>> >>>>> >> -- >>> >>>>> >> >>> >>>>> >> Kenan Wang >>> >>>>> >> Product Manager, Mobile >>> >>>>> >> Wikimedia Foundation >>> >>>>> > >>> >>>>> > >>> >>>>> > >>> >>>>> > >>> >>>>> > -- >>> >>>>> > >>> >>>>> > Kenan Wang >>> >>>>> > Product Manager, Mobile >>> >>>>> > Wikimedia Foundation >>> >>>> >>> >>>> >>> >>>> >>> >>>> >>> >>>> -- >>> >>>> >>> >>>> Kenan Wang >>> >>>> Product Manager, Mobile >>> >>>> Wikimedia Foundation >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> -- >>> >>> >>> >>> Kenan Wang >>> >>> Product Manager, Mobile >>> >>> Wikimedia Foundation >>> >> >>> >> >>> >> >>> >> >>> >> -- >>> >> Arthur Richards >>> >> Software Engineer, Mobile >>> >> [[User:Awjrichards]] >>> >> IRC: awjr >>> >> +1-415-839-6885 x6687 >>> > >>> > >>> > >>> > >>> > -- >>> > >>> > Kenan Wang >>> > Product Manager, Mobile >>> > Wikimedia Foundation >>> >> >> >> >> -- >> Arthur Richards >> Software Engineer, Mobile >> [[User:Awjrichards]] >> IRC: awjr >> +1-415-839-6885 x6687 >> >> >> > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation >
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
I reviewed all the queries used for the recent mobile A/B test analysis and they are not affected by this issue, but duplicate logging of events in MobileWebEditing should be fixed: I added it to the list of instrumentation bugs/issues for mobile that we’ve previously discussed: https://trello.com/c/5lcUPxGn
On Dec 13, 2013, at 8:11 PM, Dan Andreescu dandreescu@wikimedia.org wrote:
So Jon, Ori, and I agreed it'd be interesting to run the following queries:
duplicates in MobileWebEditing_5644223 including timestamp: about 315k, about 37k are the "success" event duplicates in MobileWebEditing_5644223 excluding timestamp: about 323k, about 39k are "success" duplicates in MobileWebEditing_6077315 including timestamp: 327, 4 "success" duplicates in MobileWebEditing_6077315 excluding timestamp: about 83k, about 3k are "success"
To get these I just grouped by the fields you guys are selecting in the query backing that graph, and got everything having count > 1. Of interest is that some things have count == 6. All the files are available in my home directory on stat1, I made them world readable. The queries were all like this:
select L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount, count(*) as duplicity from log.MobileWebEditing_6077315 L group by L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount having count(*) > 1 ;
Hm, I looked at one of these records and it seems not all the deduplications that happen because of the UNION are good. Some have different event_pageId values. So, when cleaning those tables, be careful to include that field and any others that matter.
On Fri, Dec 13, 2013 at 9:41 PM, Jon Robson jrobson@wikimedia.org wrote: So I worked out what's happening but not sure exactly why. The graph is actually more accurate now.
It seems there are duplicate rows in the database table where every column has the same value except id. So when you do a UNION as we recently started doing for the editing graphs which omits the id field you end up with non-distinct values which a UNION will throw away.
Take a look at: select * from MobileWebEditing_5644223 where timestamp = '20130806192404' and event_username = 'Bender235'
This gives two edits that were exactly the same apart from the id. Not sure why this is the case (will explore more next week)
On Fri, Dec 13, 2013 at 6:10 PM, Kenan Wang kwang@wikimedia.org wrote: Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote: Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote: I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu dandreescu@wikimedia.org wrote: I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query: https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de...
- I understand the latest graph is different from old versions of it. When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote: Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote: Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli dtaraborelli@wikimedia.org wrote: Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.org wrote: Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a drop in edits between August and September...
Is it possible that we had data issues before that were causing us to see an issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing data that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards arichards@wikimedia.org wrote:
I guess it's tough to draw a baseline considering we only have a couple months worth of data - but perhaps a good metric for our goal is returning to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that became active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org wrote:
Well, looks like IndiaSummer95 got blocked today... but I've reached out to Nick1372 and KurtWags3 via their user talk pages.
On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org wrote: > > I hope this is not too stalker-y but I think it is interesting. > > I looked briefly at the top 5 editors for both August [A] and > September [B], the most prolific editor "Indiasummer95", 520 days > registered, [1] made 2908 edits in August but only 694 in September. > The second most profilic editor "La Avatar Korra", registered for 88 > days [2] in September made 954 edits in August but only 379 in > September. Number 3, Nick1372, member for 306d days [3] had 762 edits > in August but only 445 in September, #4 was 이민혁, registered for 70 > days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 > days [5] (520 dropping to 102) > > Out of these Indiasummer95 was the most interesting so I looked at the > edits per page for August [C]: > Tell_Mama_UK 66 > Abel_Xavier 60 > Melilla 56 > Vitória_S.C. 44 > History_of_Roman_Catholicism_in_Portugal 40 > Knattspyrnufélag_Reykjavíkur 40 > Robot_Wars_(TV_series) 38 > F.C._Paços_de_Ferreira 32 > Symbols_of_Portugal 32 > Politics_of_Toledo_(Spain) 32 > > The same in September [D]: > Alfonso_I_of_Asturias 21 > Indiasummer95 13 > Vitālijs_Astafjevs 12 > A.C._Perugia_Calcio 10 > Morangos_com_Açúcar 9 > Azad_Ali 9 > Million_Muslim_March 8 > Islam_in_the_Czech_Republic 7 > Szilárd_Németh 7 > Matteo_Ferrari 6 > > Inspecting closing on the dates of his/her edits to Tell_Mama_UK (page > id 39683645) [E] > 2013-08-11 4 > 2013-08-15 56 > 2013-08-16 6 > > However if I look at the history page I don't see 56 edits from > him/her on that page (I count 36 total on the entire page): > > https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history > > So.. that makes me wonder... > 1) was something broken around 15th August in EventLogging - were we > reporting successes that weren't and thus the data is skewed (e.g. > AbuseFilter related) ? Is there a logical explanation for this or is > something indeed weird - can anyone suggest a reason? > 2) Is this people getting used to the editor/wikitext - making lots of > mistakes early on and then getting better at using it? > 3) Is it indeed seasonal / people getting bored after initial > excitement of editing? > > [1] > https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... > [2] > https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... > [3] > https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... > [4] > https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... > [5] > https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... > > [A] SELECT event_username, count(*) from MobileWebEditing_5644223 > where event_action = 'success' and timestamp LIKE '201308%' group by > event_username > [B] SELECT event_username, count(*) from MobileWebEditing_5644223 > where event_action = 'success' and timestamp LIKE '201309%' group by > event_username > [C] SELECT enwiki.page.page_title, count(*) from > MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id > = event_pageId where wiki = 'enwiki' and event_action = 'success' and > event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group by > event_pageId > [D] SELECT enwiki.page.page_title, count(*) from > MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id > = event_pageId where wiki = 'enwiki' and event_action = 'success' and > event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group by > event_pageId > [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 > where wiki = 'enwiki' and event_action = 'success' and event_userName > = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = > 39683645 group by DATE(timestamp) > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org wrote: > > I'll try to block out some time to look at creating those two graphs > > tomorrow. Let me know if you want to help. > > > > > > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org > > wrote: > >> > >> Adding Dario to this thread. > >> > >> So the main question is why is the number of edits going down from > >> august > >> to september while the number of unique editors is going up? > >> > >> Few theories: > >> > >> 1) Problem with data - seems unlikely per Jon's analysis. Also, both > >> numbers come from the same data source. > >> > >> 2) Change in the makeup of editors - not conclusive from the data > >> but it > >> seems that the number of edits from new editors and experienced > >> editors are > >> following the same general pattern. > >> > >> 3) Change in edits per session - no analysis done towards this yet. > >> > >> I'd like to take a look at: > >> > >> 1) a daily graph over the same two month period on unique daily > >> editors > >> > >> 2) a daily graph of edits per session over the same two month > >> period > >> > >> Few things to keep in mind: > >> > >> -It seems that there is not a seasonality difference in desktop > >> edits, but > >> as pointed out by Steven Walling it's not necessarily true that > >> desktop > >> edits and mobile edits will follow the same seasonality patterns. > >> > >> -Seasonality also wouldn't account for the skew between unique > >> editors and > >> raw edits. > >> > >> -This effect seems fairly centered on enwiki, however mobile captcha > >> support was introduced and may affect numbers for other projects in > >> some off > >> setting fashion > >> > >> -The spike in number of edits occurs roughly between Aug 12 - 30, > >> particularly 23-30. In that time section level editing was > >> introduced on > >> August 27. > >> > >> > >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson jrobson@wikimedia.org > >> wrote: > >>> > >>> If you look at my graphs this theory doesn't seem to hold.... > >>> > >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: > >>>> > >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk > >>>> mpinchuk@wikimedia.org > >>>> wrote: > >>>> > One theory: most of our enwiki power users tried mobile editing > >>>> > early > >>>> > on and > >>>> > made their customarily high number of edits per session in > >>>> > August, but > >>>> > they > >>>> > weren't retained on mobile and didn't stick around through > >>>> > September. > >>>> > Meanwhile, new users continue to flood in, but they only make an > >>>> > edit > >>>> > or two > >>>> > and leave. If this is the case, it makes a strong argument for > >>>> > working > >>>> > on > >>>> > things like mobile GettingStarted/KeepGoing to bump up early > >>>> > engagement for > >>>> > newbies, and article histories/contrib views and other power > >>>> > user > >>>> > features > >>>> > for oldbies :) > >>>> > >>>> This could be an interesting theory. Kenan, what do you think > >>>> about > >>>> reaching out to a handful of those editors that dropped and ask > >>>> them > >>>> why? > >>>> > >>>> --tomasz > >> > >> > >> > >> > >> -- > >> > >> Kenan Wang > >> Product Manager, Mobile > >> Wikimedia Foundation > > > > > > > > > > -- > > > > Kenan Wang > > Product Manager, Mobile > > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
It's still not clear /why/ there are duplicate events. It could be one of several things: 1) Some error in JavaScript code for mobile at some point 2) Bug in EventLogging 3) A problem occurred with backfilling data
Is uuid always unique? As if so it seems that MobileWebEditing_5644223 did indeed get corrupted some how...
These 2 numbers do not agree with one another: select count(*) from MobileWebEditing_5644223 = 1467721 select count(distinct uuid) from MobileWebEditing_5644223 = 1152600
These two are the same: select count(*) from MobileWebEditing_6077315 = 643090 select count(distinct uuid) from MobileWebEditing_6077315 =643090
On Mon, Dec 16, 2013 at 11:29 AM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
I reviewed all the queries used for the recent mobile A/B test analysis and they are not affected by this issue, but duplicate logging of events in MobileWebEditing should be fixed: I added it to the list of instrumentation bugs/issues for mobile that we’ve previously discussed: https://trello.com/c/5lcUPxGn
On Dec 13, 2013, at 8:11 PM, Dan Andreescu dandreescu@wikimedia.org wrote:
So Jon, Ori, and I agreed it'd be interesting to run the following queries:
duplicates in MobileWebEditing_5644223 including timestamp: about 315k, about 37k are the "success" event duplicates in MobileWebEditing_5644223 excluding timestamp: about 323k, about 39k are "success" duplicates in MobileWebEditing_6077315 including timestamp: 327, 4 "success" duplicates in MobileWebEditing_6077315 excluding timestamp: about 83k, about 3k are "success"
To get these I just grouped by the fields you guys are selecting in the query backing that graph, and got everything having count > 1. Of interest is that some things have count == 6. All the files are available in my home directory on stat1, I made them world readable. The queries were all like this:
select L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount, count(*) as duplicity from log.MobileWebEditing_6077315 L group by L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount having count(*) > 1 ;
Hm, I looked at one of these records and it seems not all the deduplications that happen because of the UNION are good. Some have different event_pageId values. So, when cleaning those tables, be careful to include that field and any others that matter.
On Fri, Dec 13, 2013 at 9:41 PM, Jon Robson jrobson@wikimedia.org wrote:
So I worked out what's happening but not sure exactly why. The graph is actually more accurate now.
It seems there are duplicate rows in the database table where every column has the same value except id. So when you do a UNION as we recently started doing for the editing graphs which omits the id field you end up with non-distinct values which a UNION will throw away.
Take a look at: select * from MobileWebEditing_5644223 where timestamp = '20130806192404' and event_username = 'Bender235'
This gives two edits that were exactly the same apart from the id. Not sure why this is the case (will explore more next week)
On Fri, Dec 13, 2013 at 6:10 PM, Kenan Wang kwang@wikimedia.org wrote:
Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote:
Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.orgwrote:
I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu < dandreescu@wikimedia.org> wrote:
I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query:
https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de... ?
- I understand the latest graph is different from old versions of it.
When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.orgwrote:
> Anyone know what happened here? > > > On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.orgwrote: > >> Looking at this one: >> http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful >> >> >> On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli < >> dtaraborelli@wikimedia.org> wrote: >> >>> Kenan – can you give me a pointer to the graph you are referring >>> to and confirm what data source you are using? If you’re relying on >>> EventLogging data (MobileWebEditing) the log is not affected by reverts or >>> page deletions. >>> >>> On Dec 2, 2013, at 11:53 AM, Arthur Richards < >>> arichards@wikimedia.org> wrote: >>> >>> +analytics >>> >>> >>> On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson <jrobson@wikimedia.org >>> > wrote: >>> >>>> Several theories: >>>> 1) I guess this now reflects reversions/deletions of edits? >>>> 2) We purge data collected before a certain date >>>> >>>> I suspect the former? >>>> >>>> On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org >>>> wrote: >>>> > Hey everyone. The monthly successful edits graph no longer >>>> reflects a drop >>>> > in edits between August and September... >>>> > >>>> > Is it possible that we had data issues before that were causing >>>> us to see an >>>> > issue that wasn't there? >>>> > >>>> > OR >>>> > >>>> > Is there some sort of data issue now that is causing us to be >>>> missing data >>>> > that we used to be able to see? >>>> > >>>> > >>>> > On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards < >>>> arichards@wikimedia.org> >>>> > wrote: >>>> >> >>>> >> I guess it's tough to draw a baseline considering we only have >>>> a couple >>>> >> months worth of data - but perhaps a good metric for our goal >>>> is returning >>>> >> to or exceeding our peak? >>>> >> >>>> >> >>>> >> On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang < >>>> kwang@wikimedia.org> wrote: >>>> >>> >>>> >>> Thanks to Juliusz we now have a graph of mobile editors that >>>> became >>>> >>> active editors. We can see a drop in the number of editors >>>> that became >>>> >>> active: >>>> >>> >>>> >>> >>>> http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors >>>> >>> >>>> >>> >>>> >>> On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang < >>>> kwang@wikimedia.org> wrote: >>>> >>>> >>>> >>>> Well, looks like IndiaSummer95 got blocked today... but I've >>>> reached out >>>> >>>> to Nick1372 and KurtWags3 via their user talk pages. >>>> >>>> >>>> >>>> >>>> >>>> On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson < >>>> jrobson@wikimedia.org> >>>> >>>> wrote: >>>> >>>>> >>>> >>>>> I hope this is not too stalker-y but I think it is >>>> interesting. >>>> >>>>> >>>> >>>>> I looked briefly at the top 5 editors for both August [A] >>>> and >>>> >>>>> September [B], the most prolific editor "Indiasummer95", >>>> 520 days >>>> >>>>> registered, [1] made 2908 edits in August but only 694 in >>>> September. >>>> >>>>> The second most profilic editor "La Avatar Korra", >>>> registered for 88 >>>> >>>>> days [2] in September made 954 edits in August but only 379 >>>> in >>>> >>>>> September. Number 3, Nick1372, member for 306d days [3] had >>>> 762 edits >>>> >>>>> in August but only 445 in September, #4 was 이민혁, registered >>>> for 70 >>>> >>>>> days, [4] (558 rising to 623), and #5 KurtWags3, registered >>>> for 84 >>>> >>>>> days [5] (520 dropping to 102) >>>> >>>>> >>>> >>>>> Out of these Indiasummer95 was the most interesting so I >>>> looked at the >>>> >>>>> edits per page for August [C]: >>>> >>>>> Tell_Mama_UK 66 >>>> >>>>> Abel_Xavier 60 >>>> >>>>> Melilla 56 >>>> >>>>> Vitória_S.C. 44 >>>> >>>>> History_of_Roman_Catholicism_in_Portugal 40 >>>> >>>>> Knattspyrnufélag_Reykjavíkur 40 >>>> >>>>> Robot_Wars_(TV_series) 38 >>>> >>>>> F.C._Paços_de_Ferreira 32 >>>> >>>>> Symbols_of_Portugal 32 >>>> >>>>> Politics_of_Toledo_(Spain) 32 >>>> >>>>> >>>> >>>>> The same in September [D]: >>>> >>>>> Alfonso_I_of_Asturias 21 >>>> >>>>> Indiasummer95 13 >>>> >>>>> Vitālijs_Astafjevs 12 >>>> >>>>> A.C._Perugia_Calcio 10 >>>> >>>>> Morangos_com_Açúcar 9 >>>> >>>>> Azad_Ali 9 >>>> >>>>> Million_Muslim_March 8 >>>> >>>>> Islam_in_the_Czech_Republic 7 >>>> >>>>> Szilárd_Németh 7 >>>> >>>>> Matteo_Ferrari 6 >>>> >>>>> >>>> >>>>> Inspecting closing on the dates of his/her edits to >>>> Tell_Mama_UK (page >>>> >>>>> id 39683645) [E] >>>> >>>>> 2013-08-11 4 >>>> >>>>> 2013-08-15 56 >>>> >>>>> 2013-08-16 6 >>>> >>>>> >>>> >>>>> However if I look at the history page I don't see 56 edits >>>> from >>>> >>>>> him/her on that page (I count 36 total on the entire page): >>>> >>>>> >>>> >>>>> >>>> https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history >>>> >>>>> >>>> >>>>> So.. that makes me wonder... >>>> >>>>> 1) was something broken around 15th August in EventLogging >>>> - were we >>>> >>>>> reporting successes that weren't and thus the data is >>>> skewed (e.g. >>>> >>>>> AbuseFilter related) ? Is there a logical explanation for >>>> this or is >>>> >>>>> something indeed weird - can anyone suggest a reason? >>>> >>>>> 2) Is this people getting used to the editor/wikitext - >>>> making lots of >>>> >>>>> mistakes early on and then getting better at using it? >>>> >>>>> 3) Is it indeed seasonal / people getting bored after >>>> initial >>>> >>>>> excitement of editing? >>>> >>>>> >>>> >>>>> [1] >>>> >>>>> >>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... >>>> >>>>> [2] >>>> >>>>> >>>> https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... >>>> >>>>> [3] >>>> >>>>> >>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... >>>> >>>>> [4] >>>> >>>>> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... >>>> 이민혁?mobileaction=beta >>>> >>>>> [5] >>>> >>>>> >>>> https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... >>>> >>>>> >>>> >>>>> [A] SELECT event_username, count(*) from >>>> MobileWebEditing_5644223 >>>> >>>>> where event_action = 'success' and timestamp LIKE '201308%' >>>> group by >>>> >>>>> event_username >>>> >>>>> [B] SELECT event_username, count(*) from >>>> MobileWebEditing_5644223 >>>> >>>>> where event_action = 'success' and timestamp LIKE '201309%' >>>> group by >>>> >>>>> event_username >>>> >>>>> [C] SELECT enwiki.page.page_title, count(*) from >>>> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >>>> enwiki.page.page_id >>>> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >>>> 'success' and >>>> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE >>>> '201308%' group by >>>> >>>>> event_pageId >>>> >>>>> [D] SELECT enwiki.page.page_title, count(*) from >>>> >>>>> MobileWebEditing_5644223 INNER JOIN enwiki.page on >>>> enwiki.page.page_id >>>> >>>>> = event_pageId where wiki = 'enwiki' and event_action = >>>> 'success' and >>>> >>>>> event_userName = 'Indiasummer95' and timestamp LIKE >>>> '201309%' group by >>>> >>>>> event_pageId >>>> >>>>> [E] SELECT DATE(timestamp), count(*) from >>>> MobileWebEditing_5644223 >>>> >>>>> where wiki = 'enwiki' and event_action = 'success' and >>>> event_userName >>>> >>>>> = 'Indiasummer95' and timestamp LIKE '201308%' and >>>> event_pageId = >>>> >>>>> 39683645 group by DATE(timestamp) >>>> >>>>> >>>> >>>>> >>>> >>>>> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >>>> kwang@wikimedia.org> wrote: >>>> >>>>> > I'll try to block out some time to look at creating those >>>> two graphs >>>> >>>>> > tomorrow. Let me know if you want to help. >>>> >>>>> > >>>> >>>>> > >>>> >>>>> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang < >>>> kwang@wikimedia.org> >>>> >>>>> > wrote: >>>> >>>>> >> >>>> >>>>> >> Adding Dario to this thread. >>>> >>>>> >> >>>> >>>>> >> So the main question is why is the number of edits going >>>> down from >>>> >>>>> >> august >>>> >>>>> >> to september while the number of unique editors is going >>>> up? >>>> >>>>> >> >>>> >>>>> >> Few theories: >>>> >>>>> >> >>>> >>>>> >> 1) Problem with data - seems unlikely per Jon's >>>> analysis. Also, both >>>> >>>>> >> numbers come from the same data source. >>>> >>>>> >> >>>> >>>>> >> 2) Change in the makeup of editors - not conclusive from >>>> the data >>>> >>>>> >> but it >>>> >>>>> >> seems that the number of edits from new editors and >>>> experienced >>>> >>>>> >> editors are >>>> >>>>> >> following the same general pattern. >>>> >>>>> >> >>>> >>>>> >> 3) Change in edits per session - no analysis done >>>> towards this yet. >>>> >>>>> >> >>>> >>>>> >> I'd like to take a look at: >>>> >>>>> >> >>>> >>>>> >> 1) a daily graph over the same two month period on >>>> unique daily >>>> >>>>> >> editors >>>> >>>>> >> >>>> >>>>> >> 2) a daily graph of edits per session over the same two >>>> month >>>> >>>>> >> period >>>> >>>>> >> >>>> >>>>> >> Few things to keep in mind: >>>> >>>>> >> >>>> >>>>> >> -It seems that there is not a seasonality difference in >>>> desktop >>>> >>>>> >> edits, but >>>> >>>>> >> as pointed out by Steven Walling it's not necessarily >>>> true that >>>> >>>>> >> desktop >>>> >>>>> >> edits and mobile edits will follow the same seasonality >>>> patterns. >>>> >>>>> >> >>>> >>>>> >> -Seasonality also wouldn't account for the skew between >>>> unique >>>> >>>>> >> editors and >>>> >>>>> >> raw edits. >>>> >>>>> >> >>>> >>>>> >> -This effect seems fairly centered on enwiki, however >>>> mobile captcha >>>> >>>>> >> support was introduced and may affect numbers for other >>>> projects in >>>> >>>>> >> some off >>>> >>>>> >> setting fashion >>>> >>>>> >> >>>> >>>>> >> -The spike in number of edits occurs roughly between Aug >>>> 12 - 30, >>>> >>>>> >> particularly 23-30. In that time section level editing >>>> was >>>> >>>>> >> introduced on >>>> >>>>> >> August 27. >>>> >>>>> >> >>>> >>>>> >> >>>> >>>>> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson < >>>> jrobson@wikimedia.org> >>>> >>>>> >> wrote: >>>> >>>>> >>> >>>> >>>>> >>> If you look at my graphs this theory doesn't seem to >>>> hold.... >>>> >>>>> >>> >>>> >>>>> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org >>>> wrote: >>>> >>>>> >>>> >>>> >>>>> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >>>> >>>>> >>>> mpinchuk@wikimedia.org >>>> >>>>> >>>> wrote: >>>> >>>>> >>>> > One theory: most of our enwiki power users tried >>>> mobile editing >>>> >>>>> >>>> > early >>>> >>>>> >>>> > on and >>>> >>>>> >>>> > made their customarily high number of edits per >>>> session in >>>> >>>>> >>>> > August, but >>>> >>>>> >>>> > they >>>> >>>>> >>>> > weren't retained on mobile and didn't stick around >>>> through >>>> >>>>> >>>> > September. >>>> >>>>> >>>> > Meanwhile, new users continue to flood in, but they >>>> only make an >>>> >>>>> >>>> > edit >>>> >>>>> >>>> > or two >>>> >>>>> >>>> > and leave. If this is the case, it makes a strong >>>> argument for >>>> >>>>> >>>> > working >>>> >>>>> >>>> > on >>>> >>>>> >>>> > things like mobile GettingStarted/KeepGoing to bump >>>> up early >>>> >>>>> >>>> > engagement for >>>> >>>>> >>>> > newbies, and article histories/contrib views and >>>> other power >>>> >>>>> >>>> > user >>>> >>>>> >>>> > features >>>> >>>>> >>>> > for oldbies :) >>>> >>>>> >>>> >>>> >>>>> >>>> This could be an interesting theory. Kenan, what do >>>> you think >>>> >>>>> >>>> about >>>> >>>>> >>>> reaching out to a handful of those editors that >>>> dropped and ask >>>> >>>>> >>>> them >>>> >>>>> >>>> why? >>>> >>>>> >>>> >>>> >>>>> >>>> --tomasz >>>> >>>>> >> >>>> >>>>> >> >>>> >>>>> >> >>>> >>>>> >> >>>> >>>>> >> -- >>>> >>>>> >> >>>> >>>>> >> Kenan Wang >>>> >>>>> >> Product Manager, Mobile >>>> >>>>> >> Wikimedia Foundation >>>> >>>>> > >>>> >>>>> > >>>> >>>>> > >>>> >>>>> > >>>> >>>>> > -- >>>> >>>>> > >>>> >>>>> > Kenan Wang >>>> >>>>> > Product Manager, Mobile >>>> >>>>> > Wikimedia Foundation >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> >>>> -- >>>> >>>> >>>> >>>> Kenan Wang >>>> >>>> Product Manager, Mobile >>>> >>>> Wikimedia Foundation >>>> >>> >>>> >>> >>>> >>> >>>> >>> >>>> >>> -- >>>> >>> >>>> >>> Kenan Wang >>>> >>> Product Manager, Mobile >>>> >>> Wikimedia Foundation >>>> >> >>>> >> >>>> >> >>>> >> >>>> >> -- >>>> >> Arthur Richards >>>> >> Software Engineer, Mobile >>>> >> [[User:Awjrichards]] >>>> >> IRC: awjr >>>> >> +1-415-839-6885 x6687 >>>> > >>>> > >>>> > >>>> > >>>> > -- >>>> > >>>> > Kenan Wang >>>> > Product Manager, Mobile >>>> > Wikimedia Foundation >>>> >>> >>> >>> >>> -- >>> Arthur Richards >>> Software Engineer, Mobile >>> [[User:Awjrichards]] >>> IRC: awjr >>> +1-415-839-6885 x6687 >>> >>> >>> >> >> >> -- >> >> Kenan Wang >> Product Manager, Mobile >> Wikimedia Foundation >> > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation > > _______________________________________________ > Analytics mailing list > Analytics@lists.wikimedia.org > https://lists.wikimedia.org/mailman/listinfo/analytics > >
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
yes, uuid is expected to be unique (Ori, any reason why we are not enforcing a unique key on uuid in the log db?)
On Dec 16, 2013, at 1:44 PM, Jon Robson jrobson@wikimedia.org wrote:
It's still not clear /why/ there are duplicate events. It could be one of several things:
- Some error in JavaScript code for mobile at some point
- Bug in EventLogging
- A problem occurred with backfilling data
Is uuid always unique? As if so it seems that MobileWebEditing_5644223 did indeed get corrupted some how...
These 2 numbers do not agree with one another: select count(*) from MobileWebEditing_5644223 = 1467721 select count(distinct uuid) from MobileWebEditing_5644223 = 1152600
These two are the same: select count(*) from MobileWebEditing_6077315 = 643090 select count(distinct uuid) from MobileWebEditing_6077315 =643090
On Mon, Dec 16, 2013 at 11:29 AM, Dario Taraborelli dtaraborelli@wikimedia.org wrote: I reviewed all the queries used for the recent mobile A/B test analysis and they are not affected by this issue, but duplicate logging of events in MobileWebEditing should be fixed: I added it to the list of instrumentation bugs/issues for mobile that we’ve previously discussed: https://trello.com/c/5lcUPxGn
On Dec 13, 2013, at 8:11 PM, Dan Andreescu dandreescu@wikimedia.org wrote:
So Jon, Ori, and I agreed it'd be interesting to run the following queries:
duplicates in MobileWebEditing_5644223 including timestamp: about 315k, about 37k are the "success" event duplicates in MobileWebEditing_5644223 excluding timestamp: about 323k, about 39k are "success" duplicates in MobileWebEditing_6077315 including timestamp: 327, 4 "success" duplicates in MobileWebEditing_6077315 excluding timestamp: about 83k, about 3k are "success"
To get these I just grouped by the fields you guys are selecting in the query backing that graph, and got everything having count > 1. Of interest is that some things have count == 6. All the files are available in my home directory on stat1, I made them world readable. The queries were all like this:
select L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount, count(*) as duplicity from log.MobileWebEditing_6077315 L group by L.timestamp, L.wiki, L.event_username, L.event_action, L.event_namespace, L.event_userEditCount having count(*) > 1 ;
Hm, I looked at one of these records and it seems not all the deduplications that happen because of the UNION are good. Some have different event_pageId values. So, when cleaning those tables, be careful to include that field and any others that matter.
On Fri, Dec 13, 2013 at 9:41 PM, Jon Robson jrobson@wikimedia.org wrote: So I worked out what's happening but not sure exactly why. The graph is actually more accurate now.
It seems there are duplicate rows in the database table where every column has the same value except id. So when you do a UNION as we recently started doing for the editing graphs which omits the id field you end up with non-distinct values which a UNION will throw away.
Take a look at: select * from MobileWebEditing_5644223 where timestamp = '20130806192404' and event_username = 'Bender235'
This gives two edits that were exactly the same apart from the id. Not sure why this is the case (will explore more next week)
On Fri, Dec 13, 2013 at 6:10 PM, Kenan Wang kwang@wikimedia.org wrote: Dan and I just talked and it looks like a bunch of editing data in August that we used to be seeing in the graph are not longer in that graph (it used to be roughly ~78k I think and now it's ~40k)
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote: Sorry to be clear Dan Tomasz had pointed out a 30% drop between edits in august and edits in september.
On Fri, Dec 13, 2013 at 5:22 PM, Kenan Wang kwang@wikimedia.org wrote: I think the graph is updating correctly.
I noticed the discrepancy on Nov 27th.
We only know roughly I think. You can see that Tomasz pointed out on October 7th that he noticed "Rough 30% drop month on our report card [1] for successful edits." We never explained the drop and I went back to investigate.
However if you look at the graph now, that drop is not there anymore. http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Thu, Dec 12, 2013 at 5:30 PM, Dan Andreescu dandreescu@wikimedia.org wrote: I'd like to help but it's a bit hard to follow what we know so far. Some questions:
- Is this graph updating properly now, by running this query: https://git.wikimedia.org/blob/analytics%2Flimn-mobile-data/badeecb133ac67de...
- I understand the latest graph is different from old versions of it. When was this difference noticed?
- What is the difference? Do we know accurately or just roughly?
On Wed, Dec 11, 2013 at 6:08 PM, Kenan Wang kwang@wikimedia.org wrote: Anyone know what happened here?
On Tue, Dec 3, 2013 at 1:25 PM, Kenan Wang kwang@wikimedia.org wrote: Looking at this one: http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-successful
On Tue, Dec 3, 2013 at 1:22 PM, Dario Taraborelli dtaraborelli@wikimedia.org wrote: Kenan – can you give me a pointer to the graph you are referring to and confirm what data source you are using? If you’re relying on EventLogging data (MobileWebEditing) the log is not affected by reverts or page deletions.
On Dec 2, 2013, at 11:53 AM, Arthur Richards arichards@wikimedia.org wrote:
+analytics
On Mon, Dec 2, 2013 at 11:27 AM, Jon Robson jrobson@wikimedia.org wrote: Several theories:
- I guess this now reflects reversions/deletions of edits?
- We purge data collected before a certain date
I suspect the former?
On Wed, Nov 27, 2013 at 11:26 AM, Kenan Wang kwang@wikimedia.org wrote:
Hey everyone. The monthly successful edits graph no longer reflects a drop in edits between August and September...
Is it possible that we had data issues before that were causing us to see an issue that wasn't there?
OR
Is there some sort of data issue now that is causing us to be missing data that we used to be able to see?
On Thu, Oct 17, 2013 at 3:27 PM, Arthur Richards arichards@wikimedia.org wrote:
I guess it's tough to draw a baseline considering we only have a couple months worth of data - but perhaps a good metric for our goal is returning to or exceeding our peak?
On Thu, Oct 17, 2013 at 1:52 PM, Kenan Wang kwang@wikimedia.org wrote:
Thanks to Juliusz we now have a graph of mobile editors that became active editors. We can see a drop in the number of editors that became active:
http://mobile-reportcard.wmflabs.org/graphs/edits-monthly-5plus-editors
On Tue, Oct 15, 2013 at 2:31 PM, Kenan Wang kwang@wikimedia.org wrote: > > Well, looks like IndiaSummer95 got blocked today... but I've reached out > to Nick1372 and KurtWags3 via their user talk pages. > > > On Wed, Oct 9, 2013 at 9:43 PM, Jon Robson jrobson@wikimedia.org > wrote: >> >> I hope this is not too stalker-y but I think it is interesting. >> >> I looked briefly at the top 5 editors for both August [A] and >> September [B], the most prolific editor "Indiasummer95", 520 days >> registered, [1] made 2908 edits in August but only 694 in September. >> The second most profilic editor "La Avatar Korra", registered for 88 >> days [2] in September made 954 edits in August but only 379 in >> September. Number 3, Nick1372, member for 306d days [3] had 762 edits >> in August but only 445 in September, #4 was 이민혁, registered for 70 >> days, [4] (558 rising to 623), and #5 KurtWags3, registered for 84 >> days [5] (520 dropping to 102) >> >> Out of these Indiasummer95 was the most interesting so I looked at the >> edits per page for August [C]: >> Tell_Mama_UK 66 >> Abel_Xavier 60 >> Melilla 56 >> Vitória_S.C. 44 >> History_of_Roman_Catholicism_in_Portugal 40 >> Knattspyrnufélag_Reykjavíkur 40 >> Robot_Wars_(TV_series) 38 >> F.C._Paços_de_Ferreira 32 >> Symbols_of_Portugal 32 >> Politics_of_Toledo_(Spain) 32 >> >> The same in September [D]: >> Alfonso_I_of_Asturias 21 >> Indiasummer95 13 >> Vitālijs_Astafjevs 12 >> A.C._Perugia_Calcio 10 >> Morangos_com_Açúcar 9 >> Azad_Ali 9 >> Million_Muslim_March 8 >> Islam_in_the_Czech_Republic 7 >> Szilárd_Németh 7 >> Matteo_Ferrari 6 >> >> Inspecting closing on the dates of his/her edits to Tell_Mama_UK (page >> id 39683645) [E] >> 2013-08-11 4 >> 2013-08-15 56 >> 2013-08-16 6 >> >> However if I look at the history page I don't see 56 edits from >> him/her on that page (I count 36 total on the entire page): >> >> https://en.m.wikipedia.org/w/index.php?title=Tell_Mama_UK&action=history >> >> So.. that makes me wonder... >> 1) was something broken around 15th August in EventLogging - were we >> reporting successes that weren't and thus the data is skewed (e.g. >> AbuseFilter related) ? Is there a logical explanation for this or is >> something indeed weird - can anyone suggest a reason? >> 2) Is this people getting used to the editor/wikitext - making lots of >> mistakes early on and then getting better at using it? >> 3) Is it indeed seasonal / people getting bored after initial >> excitement of editing? >> >> [1] >> https://en.m.wikipedia.org/wiki/Special:UserProfile/Indiasummer95?mobileacti... >> [2] >> https://es.m.wikipedia.org/wiki/Special:UserProfile/La%20Avatar%20Korra?mobi... >> [3] >> https://en.m.wikipedia.org/wiki/Special:UserProfile/Nick1372?mobileaction=be... >> [4] >> https://ko.m.wikipedia.org/wiki/%ED%8A%B9%EC%88%98%EA%B8%B0%EB%8A%A5:UserPro... >> [5] >> https://en.m.wikipedia.org/wiki/Special:UserProfile/KurtWags3?mobileaction=b... >> >> [A] SELECT event_username, count(*) from MobileWebEditing_5644223 >> where event_action = 'success' and timestamp LIKE '201308%' group by >> event_username >> [B] SELECT event_username, count(*) from MobileWebEditing_5644223 >> where event_action = 'success' and timestamp LIKE '201309%' group by >> event_username >> [C] SELECT enwiki.page.page_title, count(*) from >> MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id >> = event_pageId where wiki = 'enwiki' and event_action = 'success' and >> event_userName = 'Indiasummer95' and timestamp LIKE '201308%' group by >> event_pageId >> [D] SELECT enwiki.page.page_title, count(*) from >> MobileWebEditing_5644223 INNER JOIN enwiki.page on enwiki.page.page_id >> = event_pageId where wiki = 'enwiki' and event_action = 'success' and >> event_userName = 'Indiasummer95' and timestamp LIKE '201309%' group by >> event_pageId >> [E] SELECT DATE(timestamp), count(*) from MobileWebEditing_5644223 >> where wiki = 'enwiki' and event_action = 'success' and event_userName >> = 'Indiasummer95' and timestamp LIKE '201308%' and event_pageId = >> 39683645 group by DATE(timestamp) >> >> >> On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org wrote: >> > I'll try to block out some time to look at creating those two graphs >> > tomorrow. Let me know if you want to help. >> > >> > >> > On Wed, Oct 9, 2013 at 8:09 PM, Kenan Wang kwang@wikimedia.org >> > wrote: >> >> >> >> Adding Dario to this thread. >> >> >> >> So the main question is why is the number of edits going down from >> >> august >> >> to september while the number of unique editors is going up? >> >> >> >> Few theories: >> >> >> >> 1) Problem with data - seems unlikely per Jon's analysis. Also, both >> >> numbers come from the same data source. >> >> >> >> 2) Change in the makeup of editors - not conclusive from the data >> >> but it >> >> seems that the number of edits from new editors and experienced >> >> editors are >> >> following the same general pattern. >> >> >> >> 3) Change in edits per session - no analysis done towards this yet. >> >> >> >> I'd like to take a look at: >> >> >> >> 1) a daily graph over the same two month period on unique daily >> >> editors >> >> >> >> 2) a daily graph of edits per session over the same two month >> >> period >> >> >> >> Few things to keep in mind: >> >> >> >> -It seems that there is not a seasonality difference in desktop >> >> edits, but >> >> as pointed out by Steven Walling it's not necessarily true that >> >> desktop >> >> edits and mobile edits will follow the same seasonality patterns. >> >> >> >> -Seasonality also wouldn't account for the skew between unique >> >> editors and >> >> raw edits. >> >> >> >> -This effect seems fairly centered on enwiki, however mobile captcha >> >> support was introduced and may affect numbers for other projects in >> >> some off >> >> setting fashion >> >> >> >> -The spike in number of edits occurs roughly between Aug 12 - 30, >> >> particularly 23-30. In that time section level editing was >> >> introduced on >> >> August 27. >> >> >> >> >> >> On Wed, Oct 9, 2013 at 6:08 PM, Jon Robson jrobson@wikimedia.org >> >> wrote: >> >>> >> >>> If you look at my graphs this theory doesn't seem to hold.... >> >>> >> >>> On 9 Oct 2013 12:24, "Tomasz Finc" tfinc@wikimedia.org wrote: >> >>>> >> >>>> On Tue, Oct 8, 2013 at 9:32 AM, Maryana Pinchuk >> >>>> mpinchuk@wikimedia.org >> >>>> wrote: >> >>>> > One theory: most of our enwiki power users tried mobile editing >> >>>> > early >> >>>> > on and >> >>>> > made their customarily high number of edits per session in >> >>>> > August, but >> >>>> > they >> >>>> > weren't retained on mobile and didn't stick around through >> >>>> > September. >> >>>> > Meanwhile, new users continue to flood in, but they only make an >> >>>> > edit >> >>>> > or two >> >>>> > and leave. If this is the case, it makes a strong argument for >> >>>> > working >> >>>> > on >> >>>> > things like mobile GettingStarted/KeepGoing to bump up early >> >>>> > engagement for >> >>>> > newbies, and article histories/contrib views and other power >> >>>> > user >> >>>> > features >> >>>> > for oldbies :) >> >>>> >> >>>> This could be an interesting theory. Kenan, what do you think >> >>>> about >> >>>> reaching out to a handful of those editors that dropped and ask >> >>>> them >> >>>> why? >> >>>> >> >>>> --tomasz >> >> >> >> >> >> >> >> >> >> -- >> >> >> >> Kenan Wang >> >> Product Manager, Mobile >> >> Wikimedia Foundation >> > >> > >> > >> > >> > -- >> > >> > Kenan Wang >> > Product Manager, Mobile >> > Wikimedia Foundation > > > > > -- > > Kenan Wang > Product Manager, Mobile > Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
-- Arthur Richards Software Engineer, Mobile [[User:Awjrichards]] IRC: awjr +1-415-839-6885 x6687
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
Analytics mailing list Analytics@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/analytics
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
--
Kenan Wang Product Manager, Mobile Wikimedia Foundation
On Mon, Dec 16, 2013 at 8:38 PM, Dario Taraborelli < dtaraborelli@wikimedia.org> wrote:
yes, uuid is expected to be unique (Ori, any reason why we are not enforcing a unique key on uuid in the log db?)
Hm, I think a unique index would slow inserts down quite a bit. And I agree it could've caught this problem earlier but this seems like a job for more unit tests / integration tests than change in design.
On Dec 17, 2013, at 12:00 AM, Dan Andreescu dandreescu@wikimedia.org wrote:
On Mon, Dec 16, 2013 at 8:38 PM, Dario Taraborelli dtaraborelli@wikimedia.org wrote: yes, uuid is expected to be unique (Ori, any reason why we are not enforcing a unique key on uuid in the log db?)
Hm, I think a unique index would slow inserts down quite a bit. And I agree it could've caught this problem earlier but this seems like a job for more unit tests / integration tests than change in design.
right, makes sense.
Jon Robson, 14/12/2013 03:41:
It seems there are duplicate rows in the database table where every column has the same value except id.
What table? If this somehow comes from recentchanges, remember that wikidata inserts fake rows. Any and all queries against RC must filter rc_type != 5. https://bugzilla.wikimedia.org/show_bug.cgi?id=43581
Nemo
It seems there are duplicate rows in the database table where every
column has the same value except id.
What table? If this somehow comes from recentchanges, remember that wikidata inserts fake rows. Any and all queries against RC must filter rc_type != 5. https://bugzilla.wikimedia.org/show_bug.cgi?id=43581
That's a good point but no, this is an EventLogging table. So the issue is new as far as we know.