So, I got interested a few days ago in the question of how many Wikitravel contributors start editing a page but bail out before completing their edit. I started combing the Apache logs for some answers.
wikitravel.org's robots.txt hides editing pages, but there are some non-compliant spiders that still follow edit links; some even mask their identity with a fake User-Agent header.
Still, I can say with some confidence that of non-spider hits on our edit pages, only about 25% result in a "submit" post afterwards. Although I think there are a lot of people who click "edit" without the idea of seriously contributing (the aforementioned spiders; people who are curious to see what will happen; people who hit edit by mistake), this still seems quite high.
I'm wondering if anyone has similar statistics for other Mediawiki sites, other wiki-engine sites, and specifically for Wikimedia sites. I'd like to get a comparison to see what we can do on Wikitravel to cut down on these bailouts and help people finish their contributions; a rough idea of what rate of bailing out other sites get would be helpful for that.
Thanks,
~Evan
Still, I can say with some confidence that of non-spider hits on our edit pages, only about 25% result in a "submit" post afterwards. Although I think there are a lot of people who click "edit" without the idea of seriously contributing (the aforementioned spiders; people who are curious to see what will happen; people who hit edit by mistake), this still seems quite high.
No, I think it's quite plausible. I wouldn't be surprised if most people click "edit" to see what happens, and then bail out because the edit box looks completely different from the rendered article, and they think they don't have the technical knowledge to make an edit (meaning: they don't know the syntax).
Timwi
Timwi wrote:
Still, I can say with some confidence that of non-spider hits on our edit pages, only about 25% result in a "submit" post afterwards. Although I think there are a lot of people who click "edit" without the idea of seriously contributing (the aforementioned spiders; people who are curious to see what will happen; people who hit edit by mistake), this still seems quite high.
No, I think it's quite plausible. I wouldn't be surprised if most people click "edit" to see what happens, and then bail out because the edit box looks completely different from the rendered article, and they think they don't have the technical knowledge to make an edit (meaning: they don't know the syntax).
At least some of the edit hits which don't result in submits are due to work-arounds for Cache404. Sometimes if there's something wrong with a cached page I'll click edit, and then edit the URL instead of the article, eg, clicking edit for Florence puts this in the location bar:
http://wikitravel.org/wiki/en/index.php?title=Florence&action=edit
but what I really want to do is re-cache the page, so I do this:
http://wikitravel.org/wiki/en/index.php?title=Florence
-mark
Evan Prodromou wrote:
So, I got interested a few days ago in the question of how many Wikitravel contributors start editing a page but bail out before completing their edit. I started combing the Apache logs for some answers.
wikitravel.org's robots.txt hides editing pages, but there are some non-compliant spiders that still follow edit links; some even mask their identity with a fake User-Agent header.
Still, I can say with some confidence that of non-spider hits on our edit pages, only about 25% result in a "submit" post afterwards. Although I think there are a lot of people who click "edit" without the idea of seriously contributing (the aforementioned spiders; people who are curious to see what will happen; people who hit edit by mistake), this still seems quite high.
I'm wondering if anyone has similar statistics for other Mediawiki sites, other wiki-engine sites, and specifically for Wikimedia sites. I'd like to get a comparison to see what we can do on Wikitravel to cut down on these bailouts and help people finish their contributions; a rough idea of what rate of bailing out other sites get would be helpful for that.
I suspect a lot of the hits on edit pages are due to readers following red links, rather than people clicking the edit tab. I'm not sure how many hits per second Wikimedia gets, but from the profiling data I can tell you that about 8% of our backend requests (i.e. squid cache misses) are edit form requests. About 16% of edit form requests result in a save attempt.
Here's an amusing story about this phenomenon: Jerome, on first seeing the huge number of edit requests in the logs, thought we were under a DoS attack from "many IPs", and made moves to start blocking them. Luckily we set him straight before he did any damage :)
-- Tim Starling
On Wed, 2005-02-11 at 11:24 +1100, Tim Starling wrote:
I suspect a lot of the hits on edit pages are due to readers following red links, rather than people clicking the edit tab.
That's a very good point.
About 16% of edit form requests result in a save attempt.
That's the number I was looking for.
I guess there's not an easy way to find out how many people are bailing out of edits that they really want to make... except maybe to ask them. Perhaps a pop-up poll for folks who navigate away from an edit page in some way besides saving? Kind of intrusive but it might have interesting results.
~Evan
Don't think that would happen at Wikipedia. You could certainly try it at Wikitravel, though... perhaps it would be better-accepted if it only popped up for every 3rd bailout? (not per person, but total).
Mark
On 02/11/05, Evan Prodromou evan@wikitravel.org wrote:
On Wed, 2005-02-11 at 11:24 +1100, Tim Starling wrote:
I suspect a lot of the hits on edit pages are due to readers following red links, rather than people clicking the edit tab.
That's a very good point.
About 16% of edit form requests result in a save attempt.
That's the number I was looking for.
I guess there's not an easy way to find out how many people are bailing out of edits that they really want to make... except maybe to ask them. Perhaps a pop-up poll for folks who navigate away from an edit page in some way besides saving? Kind of intrusive but it might have interesting results.
~Evan
-- Evan Prodromou evan@wikitravel.org Wikitravel (http://wikitravel.org/) -- the free, complete, up-to-date and reliable world-wide travel guide _______________________________________________ Wikitech-l mailing list Wikitech-l@wikimedia.org http://mail.wikipedia.org/mailman/listinfo/wikitech-l
-- If you would like a gmail invite, please send me an e-mail. Si ud. quiere que le envíe una invitación para juntar gmail, envíeme un mensaje. Si vous voulez que je vous envoie une invitation à joindre gmail, envoyez-moi s.v.p un message. Se vce. gostaria que lhe envie um convite para juntar gmail, favor de envie-me uma mensagem. Se vuleti chi vi manu 'n invitu a uniri gmail, mandatimi n messaggiu.
On Wed, 2005-02-11 at 20:13 -0700, Mark Williamson wrote:
Don't think that would happen at Wikipedia. You could certainly try it at Wikitravel, though... perhaps it would be better-accepted if it only popped up for every 3rd bailout? (not per person, but total).
My guess was about once in every 10 bailouts, and only ask the same IP address once, and only ask if the user's never had a successful saved edit.
I think it would make a good academic paper for someone -- analyzing barriers to entry for first-time Wiki contributors. It'd be interesting to compare across large vs. small wikis, different wiki engines, different kinds of edits (red link vs. edit tab). A clever grad student could probably get some money from e.g. SocialText or JotSpot to sponsor the research.
Anywho. Might be something I pick up in my plentiful spare time.
~ESP
Evan Prodromou wrote:
I guess there's not an easy way to find out how many people are bailing out of edits that they really want to make... except maybe to ask them. Perhaps a pop-up poll for folks who navigate away from an edit page in some way besides saving? Kind of intrusive but it might have interesting results.
You could tell the difference between red link clicks and edit tab clicks by checking the referrer, or by adding an extra parameter to the URL. Detecting when the user starts typing with javascript could be another measure, although some might consider that sort of thing to be an invasion of privacy. It would certainly be an invasion of privacy to send text from the edit box to the server before the user clicks "save", unless they are warned in advance.
-- Tim Starling
On 11/4/05, Tim Starling t.starling@physics.unimelb.edu.au wrote:
It would certainly be an invasion of privacy to send text from the edit box to the server before the user clicks "save", unless they are warned in advance.
If that's true, users might need to be warned about previewing timelines and equations, which are saved even when the page is not.
http://upload.wikimedia.org/wikipedia/en/timeline/d2153e614be88c9ddb74c92670...
Angela
On Friday 04 November 2005 01:37, Tim Starling wrote:
Evan Prodromou wrote:
I guess there's not an easy way to find out how many people are bailing out of edits that they really want to make... except maybe to ask them. Perhaps a pop-up poll for folks who navigate away from an edit page in some way besides saving? Kind of intrusive but it might have interesting results.
You could tell the difference between red link clicks and edit tab clicks by checking the referrer, or by adding an extra parameter to the URL.
Even simpler solution: if the article existed when the URL was accessed it is certainly an edit, otherwise it's not (might have been attempt of creation of a new article, but we can't know that anyway).
Nikola Smolenski wrote:
On Friday 04 November 2005 01:37, Tim Starling wrote:
Evan Prodromou wrote:
I guess there's not an easy way to find out how many people are bailing out of edits that they really want to make... except maybe to ask them. Perhaps a pop-up poll for folks who navigate away from an edit page in some way besides saving? Kind of intrusive but it might have interesting results.
You could tell the difference between red link clicks and edit tab clicks by checking the referrer, or by adding an extra parameter to the URL.
Even simpler solution: if the article existed when the URL was accessed it is certainly an edit, otherwise it's not (might have been attempt of creation of a new article, but we can't know that anyway).
You can't tell that from apache logs. Checking the referrer can be done solely with the apache logs, changing the URL is a one-line change to MediaWiki. Checking whether a title exists in the database is more complicated, especially if you want to do it efficiently.
-- Tim Starling
wikitech-l@lists.wikimedia.org