Thanks for the explanations Marcel, very useful to know.

I also learned not long ago that our metrics for Opera Mini and UC browser are not representative of the real traffic those browsers drive because they server render cached versions of our sites on their own servers which they serve directly to the browsers, and they don't share their analytics.

So that 2.4% may turn out to be a much bigger percentage. But the data/analytics are kept within Opera and UC browser and we don't know about them.

It would be great if we could get more real data from those companies about the amount of users on those devices in our sites, for prioritizing accordingly support/bugs.

On Thu, Aug 10, 2017 at 7:06 PM Marcel Ruiz Forns mforns@wikimedia.org wrote:

Hi Joaquin :]

In all WMF's "browser reports", the slice classified as "Other" can mean 1 of 2 things:

*1) Ua-parser https://github.com/wikimedia/analytics-ua-parser classifies those requests as "Other".* In this case, the requests come with a UA that is not recognized by ua-parser. Probably they are, as you suggest, crawlers, bots or other kinds of uncommon traffic. This should represent the smaller part of the whole "Other" slice.

*2) The anonymization algorithm is sanitizing those requests setting them to "Other".* Browser stats data is privacy-sensitive, and we can not store it raw. There is an algorithm that sanitizes all request groups that are too uncommon, like "Opera 43 on Windows Phone", because they are so specific that they could be used to re-identify a user. Each one of those groups is really small, but all sanitized groups together make up to ~10% of the traffic.

We Analytics want to dedicate some time to this hopefully next quarter, to reduce the percentage of that slice without loosing privacy: https://phabricator.wikimedia.org/T131127

*Also*, I think the term "Other" is very confusing here, because it indicates that that big 8.8% slice is neither Safari nor Chrome nor Android nor Opera etc. But, in fact, this 8.8% includes all those browsers, it is made of all those browsers. In my opinion those requests should be labelled "Unknown" or "Sanitized" instead.

Cheers!

On Mon, Jul 24, 2017 at 2:23 PM, Joaquin Oltra Hernandez < jhernandez@wikimedia.org> wrote:

...

Thanks for sharing. This is very useful information in so many ways.

For contrast, for awareness, here is some info about the mobile site https://analytics.wikimedia.org/dashboards/browsers/#mobile -site-by-browser, which looks pretty different to desktop (last month's data):

• Safari iOS is ~40.1%
  • Mobile Safari (38%)
  • iOS Chrome (safari based) (2.1%)
• Chrome is 43.7%
  • Chrome Mobile is 42%
  • Chrome is 1.7%
• Android Browser is 2.5%, with v4 being 1.9% of it
• Opera mini is 1.3%
• UC browser is 1.1%

Nuria, do you know what is the 8.8% classified as "other"? Crawler bots?

Some highlights from the last year:

• Chrome + Safari are ~84%
• Chrome mobile surpassed Safari mobile and has kept growing, more

slowly in the last months.

• Safari mobile usage seems pretty stable and most users are on v10
• The Android browser has been steadily decreasing usage, and most of

it

...

is now on the v4 version, which means the old Android 2 browsers are less of a worry

On Mon, Jul 17, 2017 at 8:33 PM Nuria Ruiz nuria@wikimedia.org wrote:

...

Hello:

Please take a look at the new browser report with more detailed desktop site data (all wikimedia projects agreggated):

https://analytics.wikimedia.org/dashboards/browsers/#desktop

-site-by-browser

...

Some highlights:

• Data is very stable over the last year

• Chrome in the lead with 45% of traffic, closely followed by IE (18%)

and

...

FF (13%)

• The bulk of IE traffic is IE11 and IE7

• Edge shows up with 4% slowly catching up to Safari (5%)

• This data is still subject to fluctuations due to bot traffic not

identified as such. We will be working on this next year.

Thanks,

Nuria _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

---

Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

-- *Marcel Ruiz Forns* Analytics Developer Wikimedia Foundation _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

@Joaquin

I also learned not long ago that our metrics for Opera Mini and UC browser are not representative of the real traffic those browsers drive because they server render cached versions of our sites on their own servers which they serve directly to the browsers, and they don't share their analytics.

Oh, interesting. Will bring this out with the team.

@Vito

I see a huge % of "other", what does it include?

It is explained a couple emails back in this thread. Please, ping me if you can not see that, and I will paste that again.

Cheers!

On Fri, Aug 11, 2017 at 2:29 PM, Vi to vituzzu.wiki@gmail.com wrote:

I see a huge % of "other", what does it include?

Vito

2017-07-17 20:33 GMT+02:00 Nuria Ruiz nuria@wikimedia.org:

...

Hello:

Please take a look at the new browser report with more detailed desktop site data (all wikimedia projects agreggated):

https://analytics.wikimedia.org/dashboards/browsers/# desktop-site-by-browser

Some highlights:

• Data is very stable over the last year

• Chrome in the lead with 45% of traffic, closely followed by IE (18%)

and

...

FF (13%)

• The bulk of IE traffic is IE11 and IE7

• Edge shows up with 4% slowly catching up to Safari (5%)

• This data is still subject to fluctuations due to bot traffic not

identified as such. We will be working on this next year.

Thanks,

Nuria _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l

---

Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l