Talk:Usage share of web browsers/Archive 6

From Wikipedia, the free encyclopedia
Jump to: navigation, search
Archive 1 Archive 4 Archive 5 Archive 6 Archive 7


When did this page start using w3schools as a source? Unlike every other source used, w3schools is only reporting its own, single site usage statistics. That is not in any way representative of the usage share of web browsers (the topic of this wiki page). All the other sources are aggregated data from many websites.

I think w3schools should be removed from the tables, and Net Applications put back in its place. Net Applications is the most quoted source of browser market share data by major publications.

w3schools seems to have been added relatively recently, and only to the summary table and external links list. I have to wonder if it was added by the site owner for promotional purposes.

Note: Not to be confused with the W3Counter data, a legitimate aggregate data source. — Preceding unsigned comment added by (talk) 02:18, 8 July 2012 (UTC)

I added it to the external links years ago so that people who considered w3schools as representative of global web browser usage would be informed. I agree that w3schools stats should be removed from the summary table. It was probably added by someone who mistakes their site's stats for overall stats for the Internet. -- Schapel (talk) 03:33, 8 July 2012 (UTC)
Agree that w3schools is dubious in its suitability for inclusion in the summary table due to its narrow visitor market [EDIT: Ah, good stuff - see W3School already removed, I misread W3Counter as it on first glance]. Psdie (talk) 00:51, 21 August 2012 (UTC)
W3Schools is considered to be a good representative of browser usage among geek (or development-oriented) users. It can be kept as it is a widely used reference, but we have to state clearly that it is not supposed to reflect global usage. It's very specific. Dodoïste (talk) 01:02, 21 August 2012 (UTC)
IP (talk) has re-added W3Schools to summary table in revision 512656504 for some reason. I've undone - W3Schools does not represent general public, so shouldn't be in summary table. Psdie (talk) 14:07, 15 September 2012 (UTC)

Restore Net Apps stats

Are some really so desperate to dilute IE's apparent marketshare that they'll stoop to removing Net Applications, the stat provider with the largest sample base in the world?!

The unique visitors vs page views argument is utterly bogus: if anything, counting page views exaggerates the browsers used by high activity users compared with just as important low activity ones (e.g., editors + admins vs users that quickly place an order). Counting unique visits better reflects the "proportion of users" IMO; but either way, removal makes no sense.

I propose Net Applications is restored within the summary table with immediate effect. Removing a source that logs 160 *million* unique visitors a month to 40k sites, inc some of the largest on the net, is insane and should not have happened. Psdie (talk) 04:32, 20 August 2012 (UTC)

I agree we should put back NetApplications into the summary table. If it's in the historical usage share, it should be in the summary table, and vice versa. We should also make the order of the columns in all of the tables consistent, because it's confusing for IE to be in the leftmost column in some tables and Chrome in that column in other tables. I propose the order: Chrome, IE, Firefox, Safari, Opera, Other, which matches the current summary table. I think we can make those two sections even more consistent, especially with regard to the "mobile vs. desktop" distinction. It's really become difficult to update the page recently because of these inconsistencies. -- Schapel (talk) 12:55, 20 August 2012 (UTC)
Let's go forward with this idea: Change the order of columns in the Historical usage share section to Chrome, IE, Firefox, Safari, Opera, Other, which matches the current summary table. I suppose we could begin with the tables for which that is the decreasing order of usage share for the latest time period. -- Schapel (talk) 16:03, 23 September 2012 (UTC)
Good to hear Steve. As IE's marketshare genuinely changes this must of course be reflected, but it does *no one* any favours to distort the stats by removing Net Apps, regardless of fact IE is a popular hate-figure for lazy devs. IE is still *widely* used in corporate and government markets in particular. When projects like jQuery start announcing dropping of v2 support for IE8 (most recent IE till only last year!) partly because of distorted marketshare perception, WP becomes part of the problem. Psdie (talk) 01:04, 21 August 2012 (UTC)
Who is best to action restoring the Net App stats to the summary table? I don't understand why DC removed it[1] in first place when there's disagreement on the Talk page? The visitors vs page views argument is bogus - you can just as easily argue counting Page Views is misleading. Net Apps is by far largest stat source on web, omitting from summary is outright misleading. Psdie (talk) 01:44, 23 August 2012 (UTC)
Thank you for restoring, Schapel - adds some balance to the sources and now better reflects what we see day-to-day on the sites we maintain. E.g., for a popular UK secondary school as a random anecdote, IE = 48%+ of unique visitors, 55% of page views. I'd suggest NetApps is restored to previous summary tables too, but main things is current and future tables. BTW, you mention "make order of stats sites [in summary table] consistent with article". Any ideas why the sites are listed in the order they are, with StatCounter at the top? Should it not be alphabetical, or is there reasoning behind the order? -- Psdie (talk) 21:37, 25 August 2012 (UTC)
I don't know any reason for the order any more. It used to be chronological -- the stats that started at the earlier dates (the ones that provided data for a longer period of time) came first. -- Schapel (talk) 23:11, 25 August 2012 (UTC)
IP tried to sneakily remove NetApps from Summary Table again - have restored, along with entries to previous month tables, new Aug 12 table and Historical Data table. I couldn't update the non-desktop historical data - seems to need a subscription, going by links for previous data? Also, the creator of the Aug Summary Table left Clicky blank - I was going to fill in, but it's not clear how to get month data for them? --Psdie (talk) 01:58, 14 September 2012 (UTC)

Net Apps doesn't measure usage

Since Net Apps doesn't measure usage, instead it measures visitors, I suggest we remove it from this page or put big disclaimers. Should not be in summary table. From: : views is the only valid metric to look at when talking about browser usage as it directly measures how much activity (or usage) is happening on each browser. Our methodology takes account of how frequently browsers are used and also tracks multi-browser usage by individuals. It is just plain wrong to claim that "browser usage" is measured using unique visitors.

Daniel.Cardenas (talk) 20:50, 19 June 2012 (UTC)

I think this article needs some updating and clarification concerning "unique visitors" vs. "activity (or usage) happening on each browser". Both metrics are interesting to the reader, and I think the article should cover both. For example, the introductory paragraph talks about "the proportion of users" and "the proportion of visitors" suggesting to me that it is "unique visitors" which is measured. A rewrite should be considered. Also, if you want to remove sources measuring unique visitors, then W3Counter should also be removed, the article itself says that W3Counter measures unique visitors, see the Differences in measurement section. (talk) 18:46, 29 June 2012 (UTC)
Agree. There should be a summary page for measuring unique visitors. Daniel.Cardenas (talk) 00:54, 30 June 2012 (UTC)
This article starts with "The usage share of web browsers is the proportion, often expressed as a percentage, of users of all web browsers who use that particular browser". Thus we should be presenting measurements of how many users use a browser, not how much activity each browser generates. Unique visitors is a measurement of users, page views are a measurement of activity. What people coming to this page are interested in is how many people choose one browser or another (market share), not how many page views users of one browser or another generate.
The StatCounter complaint is mere propaganda to try to convince more people to quote them instead of Net Applications, despite Net Applications having a much larger sample and better methodology. Its claims are largely inaccurate, though I don't think we need to get into a point-by-point discussion of that. (talk)
The lead is changed from the original, which I wrote. With the new lead, the usage shares of browsers should total well over 100%, because many users use more than one browser. For example, if 50% use IE, 30% use Chrome, 25% use Firefox, and 10% use Safari, the total usage share for browsers would be 115%! I think actual usage share is how much each browser is actually used, and by this definition, usage share will always add up to 100%. In other words, for a particular web site, if 45% of visitors use IE, 25% use Chrome, 20% use Firefox, and 10% use Safari, then the usage share adds up to 100%, and would have to add up to 100%. This is because the same user using a different browser is counted as a different "visitor", in the same way that the same user using a different computer is counted as a different "visitor". -- Schapel (talk) 17:32, 9 July 2012 (UTC)
I think we need to take a poll. Usage is not measured by visitors but by page views and the lead should reflect this. If I visit a cnn once with chrome and 10 times with firefox what is the usage share of each browser? Daniel.Cardenas (talk) 20:08, 14 July 2012 (UTC)
There is no notion of a "visit" because browsers are stateless. All the server sees is a sequence of page requests. It has no idea how many "visits" that corresponds to. There are reasons why using page hits to determine usage can be misleading, also. Different browsers can request different numbers of documents, so a browser that has excellent caching will be undercounted in page hits, but a browser that does prefetching will be overcounted. I think if a company uses page hits as a measure of usage share, we can have a disclaimer that states that measuring usage share in this way can lead to statistics that are biased. The article already has a general disclaimer to this effect, but we can add one to each table that includes statistics in such a way. Of course, all these stats are biased one way or another anyway -- they only differ in which browsers they are biased towards or against and by how much. -- Schapel (talk) 15:34, 15 July 2012 (UTC)
I don't understand stateless, because the state is maintained in the cookie. A visit is tracked by seeing if it has the same cookie and only registering one visit regardless of number of page requests. The server sees many things and tracks many things beside page requests, such as visits by users, to and from pages, etc... Daniel.Cardenas (talk) 19:44, 15 July 2012 (UTC)
Cookies exist to give browsers state. But even with a cookie, you cannot distinguish between multiple pages views during one "visit" and multiple "visits", because cookies persist between "visits". There is truly no such thing as a "visit" to a web server, so trying to define usage share in terms of "visits" is futile. -- Schapel (talk) 21:40, 15 July 2012 (UTC)
I deleted the visitor wording from the intro, but it looks messed up to me. I'm thinking the second sentence should be deleted. What do you think? Daniel.Cardenas (talk) 17:11, 16 July 2012 (UTC)
Why remove the mention of visitors? Visitors can be easy to detect, and I think it's the metric used by most stats. It's the number of "visits" made by visitors that is hard to measure, because there is no notion of "leaving" a website in general. -- Schapel (talk) 15:02, 17 July 2012 (UTC)
I've changed the definition back to the original wording I used years ago, which was in the article for most of that time. If anyone sees a way it could be improved, we can certainly discuss changes. It's actually carefully worded, and if it's changed even slightly, it may not correspond to the way the data in the article has been collected. -- Schapel (talk) 14:23, 19 July 2012 (UTC)
I am sorry, but it's hard for me to follow the reasoning for using the word "visitor" without any further explanation. My understanding is that Statcounter, Wikimedia, and others count page views, aka. page requests, whereas Net Applications and W3Counter count unique visitors. Shouldn't the lead be clear about this? The lead could say something like this instead:
One definition of the usage share of web browsers is the proportion of page views counted on a group of websites. Another definition of the usage share of web browsers is the proportion of unique visitors counted on a group of websites. Web browser usage share varies from region to region as well as through time. -- (talk) 01:58, 21 July 2012 (UTC)
The only stats site I know that uses page requests is StatCounter. Wikimedia apparently also uses Apache server logs, which have one entry per page request, but it isn't a stats site. The explanation given by StatCounter makes it pretty clear that they're doing something unusual, because they explain why they aren't using the usual method of counting unique visitors. The article has a relatively long, detailed section on page requests, so I think this method is covered more than adequately already. -- Schapel (talk) 01:33, 23 July 2012 (UTC)
Thanks for pointing out that counting 'raw' page views are relevant only to StatCounter and Wikimedia - I was not aware of that. I still think the lead has issues, though. The core of the problem lies with how you define web browser usage share. What is the definition? It is confusing to read that web browser usage share is "the proportion (...) of visitors ..." and then learn that StatCounter, a key stats site, does not measure unique visitors. (talk) 14:43, 23 July 2012 (UTC)
As far as I can tell, StatCounter does not refer to their statistics as usage share. The StatCounter section of the article also does not refer to their stats as usage share. It's only in the summary table, which says it lists usage share and lists stats from StatCounter that there is a connection between the term and StatCounter data, as least as far as I can see. I suppose it could be a bit confusing, but will most readers ever think through in that much detail? I think the fact that the whole usage share vs. page hits distinction has only become an issue recently means that most readers just don't pay that much attention to notice. I suppose we could add a sentence to the beginning of the Accuracy section that states that most statistics sites measure usage share, but StatCounter uses page hits. -- Schapel (talk) 16:04, 23 July 2012 (UTC)
I see it all over the place. Examples:
Daniel.Cardenas (talk) 15:53, 28 July 2012 (UTC)
Ah, I didn't see those. I checked the FAQ and the pages that show their stats, and they don't use that term at all. -- Schapel (talk) 23:39, 28 July 2012 (UTC)
Has any of the authors of this page checked on the claim that Net Applications weights business-hours usage? (talk) 03:57, 21 September 2012 (UTC)

Updated graphs?

When will the two StatCounter graphs be updated the next time (Countries by most used web browser and Usage share of web browsers (Source StatCounter))?

The differences aren't that huge ([1], [2]) but at least things like Sweden have also moved to Chrome, making it more clearly pass the falling IE graph. --Diblidabliduu (talk) 14:14, 10 October 2012 (UTC)

Feel free to update them. If you want help let me know. Also deleting them because they are stale tends to motivate someone to update them. Daniel.Cardenas (talk) 16:55, 10 October 2012 (UTC)

Wikimedia bar chart

I think we need to agree on that Wikimedia bar chart, because it looks ugly. There are a few problems:
1. Internet Explorer ruins the layout, so I preferred to put I.E. and we can mention somewhere that I.E. = Internet Explorer.
2. If we decide to put the full name, we should make the chart somewhat symmetrical, by having equal spaces between the browsers, which would not be very difficult, but it would still look ugly and a bit confusing, you can look at the history.
3. We need to agree on the size of that chart, I don't see the problem with it being big, as there is enough space there. I've put the bars two times bigger than the actual numbers for better view, someone increased it to three times. While this is not a problem, if a browser exceeds 33%, it will be necessary to shrink it to 2x. This is why I started with 2x and not 3x. Anyway, this can be adjusted when the time comes.
4. The size of the spaces between the bars should be agreed upon, although 5 pixels seems good to me. A3e6u9 (talk) 08:34, 12 October 2012 (UTC)

A3e6u9 - I don't understand your reasoning for reverting my improvements to the chart, which I spent considerable time on. By reverting the caption to spanning THREE lines instead of one (including unnecessary "I.E. = Internet Explorer" footnote when we've always just used the full name), the chart no longer fits within the top section and overlaps the Accuracy heading. I request that you restore my edits to the caption - it just looks clunky now. -- Psdie (talk) 09:12, 12 October 2012 (UTC)
Just noticed you've also increased the size of the chart again. I don't think you're considering users with smaller screens - I reduced the size to accommodate mobile etc users at ~800px. There's zero need for it to be so big - it doesn't improve legibility, just runs into the TOC. -- Psdie (talk) 09:19, 12 October 2012 (UTC)
I've put your 3x again, although it might not last long with the increase in popularity of Chrome, another 4% and it will have to revert to 2x. I'd agree with the full name if it did not ruin the chart. I don't think there is a way to put the whole name in one row. Anyway, I think we need some other opinions, as we seem to disagree on this one. I will put your size back for mobile readers, I did not know that. If it will have to go back to 2x it will look bad with your size. Anything else? A3e6u9 (talk) 09:55, 12 October 2012 (UTC)
Thanks for restoring size. Yep, if Chrome increases above 33%, we'd need adjust the scale - but good for now. You say it's not possible to put full Internet Explorer name in the chart, instead of "I.E." - but that's exactly what it's been until your edit. Your removing my "Sep 2012" abbreviation also prevents caption fitting on single line. -- Psdie (talk) 17:49, 12 October 2012 (UTC)
I know what has been and it was ugly with the full name. My inspiration when I created this bar chart was this: but please don't "improve" that, as it seems very good to me. Is there a rule for captions on a single line? If there is, the first caption was the shortest. A3e6u9 (talk) 05:59, 13 October 2012 (UTC)
Changing the caption to 3 lines instead of 1 not only looks ugly, but at most screen sizes causes the graphs to unnecessarily overlap the next content section ("Accuracy"). I didn't make the edits arbitrarily, I made them to improve layout. Why you decided to revert, IDK, but I'm not going to start an edit war. -- Psdie (talk) 01:42, 15 October 2012 (UTC)
My personal opinion is that is I.E. is better. Less clutter and establishes an abbreviation that can be used in the rest of the article. Daniel.Cardenas (talk) 02:24, 15 October 2012 (UTC)
A3e6u9 has now reduced caption from 3 lines to 2, meaning doesn't overlap next section - so I'm happy enough. Still not personally a fan of the abbreviation, but it's now 2 against 1 :) -- Psdie (talk) 10:57, 15 October 2012 (UTC)

Vandalism from Saudi Telecom Co user

IP / has re-added W3Schools to summary table, despite consensus here and acknowledgement from the provider that they do not represent the general public (only narrow web developer market). Further they have remove Net Applications from the summary table (present here since the article was first added) with no justification whatsoever. Warning has been posted on their talk page and I propose reporting them should they continue to vandalise the page. -- Psdie (talk) 23:25, 15 September 2012 (UTC)

This individual is now regularly vandalising the page. They've switched to IP (talk, but clearly same user as both are Saudi Telecom Co DSL IPs and both removing Net Apps from the summary table each time. Thank you to Jasper Deng for fixing. -- Psdie (talk) 02:07, 17 September 2012 (UTC)

For reference, here's the STC IPs used so far:,,,, Hopefully though they've got bored of the game now, or are ready to discuss their rationale here on the Talk page. Thanks -- Psdie (talk) 15:27, 17 September 2012 (UTC)

Seems not - they've just reverted Jasper's edit too. It's a pity, but I suspect temporary protection from anon edits may be necessary? -- Psdie (talk) 02:34, 18 September 2012 (UTC)

 Done Protected for a week.--Salix (talk): 08:08, 18 September 2012 (UTC)

Thanks Salix! Can hopefully get some peace at last -- Psdie (talk) 13:37, 19 September 2012 (UTC)
Unfortunately this sad individual is back, removing Net Apps without reason or discussion. Saudi IPs again (, Nothing better to do? -- Psdie (talk) 14:50, 1 October 2012 (UTC)
Permanent semi-protection against anon IP edits requested. -- Psdie (talk) 15:22, 1 October 2012 (UTC)
I hope there is a better alternative. We have gotten good edits from anonymous people in the past. Daniel.Cardenas (talk) 16:16, 1 October 2012 (UTC)
Hi Daniel - I can't think of any scenario when someone would have something valuable to contribute, but couldn't take 2-4 one-off minutes to sign up for an account? More generally, personally I'd have no objection to the whole of WP being protected against non logged-in edits - it means the user's edits are traceable across WP so we don't get trolls like this one exploiting dynamic IP addresses. No problem at all with login accounts being anonymous (no name etc) - that's not the issue. It's the ease of defacing from dynamic IPs, without traceability. Interested in others' views though, as a topic of general interest for WP. -- Psdie (talk) 16:02, 10 October 2012 (UTC)
Hi Psdie - If we hadn't blocked anonymous users we would have had the summary table updated by now. Although I agree it was better to block the vandalism. Daniel.Cardenas (talk) 18:49, 20 October 2012 (UTC)
I'd argue that anon user would have just have to spend 3 mins (less?) registering, then would be encouraged to further participate through update notifications etc. But fair enough that very occasionally there'd be a potentially useful contributor put off by laziness. Less of a sensitive subject at moment now the Saudi guy has calmed down :) -- Psdie (talk) 16:16, 22 October 2012 (UTC)

Wikimedia usage numbers

Looking at the sources of the Wikimedia stats, the numbers for Chrome can be seen in the squid reports, but the MSIE and Firefox numbers are different. This is true for at least a few months, as far as I checked. Where does this discrepancy come from? — Preceding unsigned comment added by (talk) 14:33, 8 November 2012 (UTC)

The Chrome numbers only come from non-mobile, the other browsers have mobile numbers, so it's non-mobile + mobile. As far as I know, Chrome is used on Android also, so I don't know why it does not appear in the mobile section.A3e6u9 (talk) 07:00, 12 November 2012 (UTC)

Proposed template and map for regional variations

It looks messy now. I propose list 1 or 2 continents where the browser if most popular plus perhaps the country where it is most popular. What do you think? Daniel.Cardenas (talk) 19:45, 10 November 2012 (UTC)

Well, I agree for that. It would be better for me (browser per continent) cause it's not easy for to check all the realms or dependencies.--Peeperman (talk) 01:18, 12 November 2012 (UTC)
Hi. I think showing continent stats neither as a template nor as a map is misleading. I am against it. I know statcounter has continent stats, however, I think the map of continents is misleading. Asia, for instance, big continent, you can't simply paint it all green. According to WP:CON, I'm removing the second map until there is a consensus. --Free ottoman (talk) 13:40, 13 November 2012 (UTC)
I think that continent stats are missing the point of regional variations, because that's what the map is trying to track. If it's easier, just avoid the really small countries, like vatican, san marino, monaco, etc. Still on the subject, I think it's good that you've removed Antarctica, too few people to be representative in any way. A3e6u9 (talk) 14:06, 15 November 2012 (UTC)
I was talking about the text, not map or even stats. Better than listing a bunch of countries, because it is less meaningful then continent. Daniel.Cardenas (talk) 03:52, 17 November 2012 (UTC)
If you want to propose the text, please open this in a new chapter. About the current subject we already have here two people who reject the idea of showing continent stats. According to WP:CON, the proposal arguments are too weak to be put into the article. Thank you.--Free ottoman (talk) 18:47, 15 November 2012 (UTC)

Wikimedia bar chart

There are two questions regarding the chart, I am creating separate sections, so that hopefully we can fix the problems with the current chart without waiting for a conclusion regarding which chart should appear. Monty845 08:05, 23 November 2012 (UTC)

I've notified the two primary parties to the edit war of this discussion: User:Free ottoman on their talk page and User:A3e6u9 in the AN/I thread they started. Monty845 08:21, 23 November 2012 (UTC)

The current chart of non-mobile browsers copies the data from [1], and uses the percentages provided there. However the percentages provided there are based on the total of all browsers, not just the non-mobile ones. There were 142,357M requests in October classified as "Browsers, non mobile", the top 5 browsers had 52,954m, 37,261m, 32,992m, 9,662m, and 5,299m. This works out to percentages of 37.20%, 26.17%, 23.18%, 06.79%, and 03.72%. Pending a conclusion to the second question, the chart should be updated to reflect this, and the bar lengths adjusted accordingly. Before making an edit request, I want to give everyone a chance to object to my proposed course of actions, but I think it should be uncontroversial pending the outcome to the broader question presented below. Monty845 08:05, 23 November 2012 (UTC)

There is a question about whether including mobile in the summary at the top of the page is misleading. Is it misleading, and if so, is the current version, which includes a non-mobile browser chart the appropriate solution? Monty845 08:05, 23 November 2012 (UTC)

I don't care if the chart includes mobile or not, I just want it to be correct. Is this a matter of debate? Should we debate whether we want a correct chart or an incorrect chart? My chart was correct, although it included the mobile stats. Can someone dispute the fact that it was correct? I can easily show that the chart of user free ottoman is incorrect, as 20% was missing and 3x3 does not equal 13. Also, if the chart included mobile usage, is this sufficient reason for it to be deleted? I think not. A3e6u9 (talk) 08:23, 23 November 2012 (UTC)
Why do you think the chart is not correct. What is wrong with it?--Salix (talk): 11:53, 23 November 2012 (UTC)

A3e6u9 was reported for violating WP:HA, WP:CON, and WP:3RR. He even called me a little girl on this board. After he didn't respond to debate the issue on article talk page , A3e6u9 said he doesn't care about the chart anymore. By the way, the chart has reference from wikimedia statistics.Free ottoman (talk) 09:04, 23 November 2012 (UTC)

Your personal spat with the other user is NOT the foray of this discussion, and is not germane at all to it. If you're going to rehash those, then you're just as bad as A3e6u9 and his "I want it fixed, I'm just not going to tell you how I think it needs to be fixed" attitude. Neither of you will be around for long if it continues (✉→BWilkins←✎) 10:30, 23 November 2012 (UTC)
It was correct before user free ottoman "fixed it". If he wants to remove the mobile stats, he needs to recalculate and make it correct. Supplanting the correct numbers with incorrect numbers is not acceptable. Deleting the chart because it included mobile stats seems silly to me, but he has removed almost every chart on the page. I must quote another user here: "15:12, 13 November 2012‎ Palacesblowlittle (talk | contribs)‎ . . (140,765 bytes) (+40)‎ . . (Undid revision 522822374 by Free ottoman (talk)there is no anything sub optimal in including, also there is no opposition, please stop the overweening deletion.)". Also, please put back this chart: someone has worked for it and it has no issues. A3e6u9 (talk) 13:03, 23 November 2012 (UTC)

He opened a duplicate ticket on administrator's board, whereas in the original ticket, you told us to discuss the issue here.--Free ottoman (talk) 11:10, 23 November 2012 (UTC)

Give me proposal, we discuss about it. And seriously, why does he think the chart is not correct? What is wrong with it?--Free ottoman (talk) 11:09, 23 November 2012 (UTC)

What is wrong is that the individual percentages don't total to 100%. How can we alter the chart so that a total of 100% is reached? Kim Dent-Brown (Talk) 12:29, 23 November 2012 (UTC)
Okay do it, put all the other to make it total %100.--Free ottoman (talk) 12:38, 23 November 2012 (UTC)
Or make a single "other" column, from the figures above it seems this is about 3%. --Salix (talk): 12:41, 23 November 2012 (UTC)
Okay, put a new "other" column. nice.--Free ottoman (talk) 12:45, 23 November 2012 (UTC)
OK, how's this? Kim Dent-Brown (Talk) 12:59, 23 November 2012 (UTC)
That's even worse, if it was possible. If you are not capable of correcting it, don't bother. I've explained a few times what is wrong with the current chart, I will not repeat yet again. A3e6u9 (talk) 13:22, 23 November 2012 (UTC)
OK consider this a formal warning. If your not capable of refining your language you will be block on the next insult or sarcastic comment. Yes there are problems with the current chart but you can refere to this without the language. Just explain simply why it is wrong, you may need to repeat yourself. --Salix (talk): 13:32, 23 November 2012 (UTC)
It's incorrect because: 1.the numbers don't add to 100%, it needs recalculation if the mobile stats are removed; 2.all the numbers need to be multiplied by two, to save space, they used to be multiplied by three, but after recalculation Chrome will be over 33%, and will exceed the size of the chart if multiplied by three. 3.after multiplication, the numbers are all wrong, all except Chrome, but maybe some of us have not passed kindergarten. 4.the "other" section has exactly the same colour as "Safari", which is misleading. 5.there are probably other reasons, but I won't bother. I'd suggest putting back the old numbers, as they are correct and will be easily maintained. Recalculations are not easily done, considering that we've seen errors in multiplications, so next month the problem will appear again. Just include the mobile stats and spare us the headaches. A3e6u9 (talk) 13:45, 23 November 2012 (UTC)
Thank you for the civility.--Salix (talk): 14:21, 23 November 2012 (UTC)
And where are you sources that lead you to the 2x and 3x numbers? This is your chance to make policy-based discussion points with dozens of admins watching. If you're not going to bother making and argument valid points, then don't bother editing the article again, and save yourself the pain of trying to work within the way this community works. (Someone in their right mind please tell me why this is updated monthly, when Wikipedia is WP:NOTNEWS?) (✉→BWilkins←✎) 14:04, 23 November 2012 (UTC)
So we have two options, 1) use stats for just the non-mobile browsers, with an others column and scaled to add to 100%. 2) use stats for all browsers with columns for mobile browsers and others. With option 1) we have 37.20%, 26.17%, 23.18%, 06.79%, and 03.72% and

2.94% for others. For option 2) the main browsers have 30.60%, 21.53%, 19.07%, 5.58%, 3.06%, other non-mobile 2.46%, all mobile 17.62%. A further refinement of option 2) would be to list the biggest mobile browsers, but that does make the table bigger.

I see A3e6u9 point about multiplying the percentages by a constant to get the bar length. I don't see scaling the visual display as a conflicting WP:RS, routine calculations are allowable by WP:CALC. It might be worth adding a comment in the text or an edit notice to describe the methodology used to get the table. --Salix (talk): 14:21, 23 November 2012 (UTC)
Something like this? "Note: bar lengths below are each THREE times the actual percentage share, in order to make difference more visible and avoid wasted space." It's already there. A3e6u9 (talk) 14:29, 23 November 2012 (UTC)
Administrators: Can you please change the color for "other" from pink to light-grey? And could you please remove the word "only" in the template?--Free ottoman (talk) 14:57, 23 November 2012 (UTC)
Thanks!--Free ottoman (talk) 15:07, 23 November 2012 (UTC)
Salix, I tip my hat to you, the chart is perfect. One minor detail is that I'd like "other" to be of this colour: #808080 (it's a darker grey), to be more visible. Plus, I was talking about restoring a chart, because the author has probably worked for it and it would be a shame not to post it there. I'm talking about this: A3e6u9 (talk) 15:46, 23 November 2012 (UTC)
Looks like the data of the chart has been recently updated. It seems like a good addition to the article showing the long term trends.--Salix (talk): 21:16, 23 November 2012 (UTC)

Wikimedia: All requests VS HTML pages

Right now for the table showing "Usage share data from Wikimedia visitor log analysis report", we are basically showing the data from the "All requests" column rather than the "HTML pages" column from the Wikimedia visitor log analysis report.

I propose we switch to using data from the HTML pages column as this is more representative of actual users visiting the Wikimedia sites. If you look at the log, pretty much for all the major browsers the percentage figures in All requests is slightly higher/lower than the percentage figures in HTML pages. All except Google Chrome, which is roughly 5% higher (as of May 2012). Well why is this? I strongly suspect this due too Google Chrome's aggressive use of its pre-rendering and pre-loading features. In fact this is such an issue for accurately representing Browser usage share that Globalstats Statscounter have as of April 2012 adjusted their Browsers stats to remove the effect of pre-rendering on their stats[2].

We should find out the difference between html page and not. Right now I see this page is .php. Does that count as html? Daniel.Cardenas (talk) 14:56, 23 May 2012 (UTC)
PHP was originally an overlay/add-on to HTML. Now some pages are almost 100% PHP, but PHP's end output is HTML, so I think yes PHP is the same as HTML. :- ) Don 15:51, 27 May 2012 (UTC)
What PHP or HTML is is irrelevant here, the important thing is what the columns so named represent. All probably includes images, style sheets and javascript files, while viewed articles probably is what is interesting (the .php is part of query strings when editing or requesting special actions). --LPfi (talk) 11:06, 28 May 2012 (UTC) Do you think watchlist page counts as HTML then? So a browser that is higher in HTML pages means that it caches images more or is it that the pages viewed have fewer images, style sheets, etc... Daniel.Cardenas (talk) 17:03, 30 May 2012 (UTC)
All content page loads count in both. All request however could include AJAX calls and partial page refreshes including other non-page load actions, things like App activity that isn't browser universal. If I had to take a partially informed guess that last part is what is probably accounting for the usage difference in the % data-sets since MEIE doesn't have large app support in the same way Firefox and Chrome do. Either way HTML calls are actual web page requests. --Karekwords?! 04:43, 1 June 2012 (UTC)
I sent an email to Erik, who creates the statistics, requesting he comment here. Daniel.Cardenas (talk) 17:11, 1 June 2012 (UTC)
Ok that should help. Also I forgot to mention that Chrome, Opera, and Safari all pre-render content which can have an impact on these types of stats. --Karekwords?! 19:50, 1 June 2012 (UTC)
No response from Erik, so going with HTML page views seems like the best thing to do. My concern is that content scrapers will often use I.E. as their user agent string.   Daniel.Cardenas (talk) 02:13, 20 June 2012 (UTC)

I added HTML page version of wikimedia stats. Daniel.Cardenas (talk) 15:52, 2 December 2012 (UTC)

Wikimedia usage

It looks like the Wikimedia usage reports haven't been updated in a few months, since the reports haven't been generated for any month after last October. Does anyone know who usually generates these reports and why they haven't been made? It would be nice to see the development of browser usage on Wikimedia sites. (talk) 15:50, 12 February 2013 (UTC)

The reports have been updated. Something is strange about January, though. FireFox dropped from 19.07% to 15.55% compared to the previous month. Safari dropped from 5.63% to 2.55%. However, iPad and iPhone now show up as new items, under desktop for some reason. Maybe they request the desktop site and the browser changes the user agent string? The same change applies to the tablet and other mobile sections. It's also of note that Chrome doesn't show up anywhere but under desktop browsers, even though it is the standard browser on Nexus tablets and should at least show up in there. Could it be that it is counted as a desktop browser in all cases? (talk) 16:27, 27 March 2013 (UTC)
Slightly changing subjects:
Per previous discussion we should report HTML usage, not all requests.
Daniel.Cardenas (talk) 17:00, 28 March 2013 (UTC)

Overestimation of chrome's usage due to pre-rendering

It is really misleading to state "some browsers (including google chrome)" perform "prerendering", consider chrome is the only browser listed in the table that actually does this. The statement made in the article should carry clear meaning and not leaving misconception to form in reader's mind.

If one were to argue this impact isn't significant, he or she should edit the sentence to state such, although anything more than 1% should be considered significant. Not to mention, other vendor has claimed pre-rendering in chrome overestimate the usage share by more than 4%. Such significance cannot be left out.

The top sites, speed dial function are present in opera, safari, chrome, firefox and IE, they are not technically pre-rendering, and it occurs at a much low frequency to very limited website, opera's speed dial almost never refresh, firefox's top sites only grab the image after user initiated visit, Safari's low marketshare makes that impact even less noticable, IE has no preview whatsoever. Nobody argue that these behavior affect usage count in any significant way, which is in contrast to the fact that net applications, statcounter, and google itself all stated chrome's pre-rendering over estimate the usage.

The explanation given by statcounter sounds like, because we can't exclude A (a much minor impact factor) due to lack of data, we will exclude B (a much higher impact factor) even if we have the data, that is not a correct attitude in providing accurate data as it possibly can. — Preceding unsigned comment added by (talk) 02:10, 20 February 2013 (UTC)

I now have an account, so let me sign thisWaynelevin (talk) 14:33, 20 February 2013 (UTC)

The reference states that other browsers do prerendering. Be aware of wp:3RR. Daniel.Cardenas (talk) 18:35, 20 February 2013 (UTC)
Chrome, Safari and Opera are the browsers mentioned in the ref. I have edited the article text to reflect that. As talk says, it needs to be specific. Useerup (talk) 06:41, 22 February 2013 (UTC)

The problem is that, as currently written, we are taking statcounter's words for granted. let me repeat, Opera's speed dial thumbnail doesn't seem to update at all, safari has such a small market, the preview would definitely not contribute significantly to the usage determination. I can understand statcounter making excuse for themselves, but we need to somehow let readers be aware of that. any suggestions? Waynelevin (talk) 18:26, 1 March 2013 (UTC)

Statcounter took out the pre-rending chrome but chrome's usage didn't drop so it wasn't making a difference ( If it bothers you that much, do research about how statecounter is excluding this data. (talk) 20:12, 13 March 2013 (UTC)

map with new file size

Freenode IRC wikimedia admin told me I should upload a new file for different file sizes, so I made a new map with bigger file size to properly describe the European microstates.--Free ottoman (talk) 12:18, 2 March 2013 (UTC)

Statcounter chart

Statcounter chart moved under Statcounter company column.--Free ottoman (talk) 12:20, 2 March 2013 (UTC)

Tables sorting

The tables don't sort correctly when you sort them by date. They get sorted lexicographically instead of chronologically. — Preceding unsigned comment added by (talk) 12:35, 12 May 2013 (UTC)

Discussions related to the country/browser map

Discussions related to the map are only under talk of the map to prevent duplicate talk page.--Free ottoman (talk) 00:21, 5 February 2013 (UTC)

Considering there are differing oppinions on the best way to derive these statistics, it would only be fair to include the same graphic of usage for the two types of information gathering.

This article seems to mention Google Chrome a lot more than it should suggesting that the author's bias is Chrome. Considering the only scientific method of measuring statistics here is putting IE at 30+% above the rest, you would assume that to be the most prevalently used browser (aka more people use the browser, rather than more pages are used by less instances of the browser. — Preceding unsigned comment added by Dawesi (talkcontribs) 23:08, 2 June 2013 (UTC)


I would just like to express my thanks to those who keep this useful article up to date. —Anne Delong (talk) 14:23, 17 June 2013 (UTC)

I think so. Manzzzz(talk) 12:51, 4 August 2013 (UTC)

Wikimedia data

I would like to know why wikimedia stats stoped in october 2012 and come back just in april 2013?--Marceloml (talk) 22:56, 8 July 2013 (UTC).

I haven't analysed carefully, but is it because people took to 'improving' the Wikimedia figures by adding and re-arranging them, which made doing a monthly update a lengthy and complicated business? --Nigelj (talk) 16:34, 2 August 2013 (UTC)
I think it's a common problem on Wikipedia in general and this article in particular. People keep wanting to add more and more info, but especially on a page that needs to be kept updated regularly, it eventually becomes too much work and it starts breaking down. Even if editors can keep up, it can make articles long, boring, and not particularly useful. I suggest simplifying a bit. Does anyone in the world really need all the most recent months' stats collected into this one article? It just makes the tables lengthy and hides the overall story -- it's hard to see the forest for the trees in other words. Why not present quarterly stats -- it would make nearly all the tables fit on one screen so they would be easier to read, and it would be much less work to update the article. If someone desparately needs the latest months' stats from one source or another (although I can't see why it would be important) they can just go to the sources. -- Schapel (talk) 22:47, 2 August 2013 (UTC)
Alright, I suppose I'll change the tables to quarterly as there are no objections. -- Schapel (talk) 15:55, 8 September 2013 (UTC)
Good move; I'd never suggest deleting info, but this is an important page, so it needs to maintain clarity. At some stage, someone will have to consider how to move some of it to a secondary page. Heenan73 (talk) 11:31, 11 October 2013 (UTC)
I suggest you keep the monthly statistics. Sure, looking at an entire table, that may look unwieldy, but looking at one series of data it will be easier to see at a glance if there is a yearly or seasonal trend. Also, if you want to compute a trend given the data, your input will hopefully be better, given monthly data. cheers/Rolf185.26.182.35 (talk) 23:59, 30 November 2013 (UTC)

Regarding Weighting and Net Applications

I made some changes to the methodology summary section in an attempt to more accurately reflect what Statistical Weighting actually is and why it is done. Previously the section referred to weighting as an introduction of bias and claimed it was a form of Sampling Bias which is simply incorrect. Statistical Weighting is done specifically to combat Sampling Bias, particularly when there's a known and measurable level of Sampling Bias, which is the case in Web Browser traffic to any given website when measured by local driving the traffic. There have been a number of other revisions to that section that raise some concern, in my mind, that the article is being edited with a non-neutral view towards this particular topic and, in specific the company it was addressing, in addition to there being some confusion on statistical methodology and terms. Figured if anyone is feeling particularly verbose on this topic it'd be better to discuss it here and develop a neutrally driven consensus over disputes in the section instead of individual driven edits. --Karekwords?! 10:00, 1 January 2014 (UTC)


I think that they should be removed from the stats section. Here is a quote from there FAQ. "These sample reports are intended to demonstrate the format and capability of Net Market Share reports. The data within these sample reports is either out of date or fabricated." (talk) 15:10, 24 June 2013 (UTC)

If that's true, I agree. Feel free to remove the bogus information and move NetApplications to the Older Reports. -- Schapel (talk) 09:17, 26 June 2013 (UTC)

That sentence must be in reference to the sample reports in the FAQ page itself, like when one graphic shows 69.54% to iPhone, and 5.02% to Android. A tad outdated. While the highly dubious scaling according to somewhat arbitrary numbers for "internet population" taint the numbers, the real numbers are most probably neither outdated nor outright fabricated. (talk) 13:09, 26 June 2013 (UTC)

Source for faking

Netapplications statistics do run against common sense, with MSIE market shares that are double those of any other independent source, including wikipedia itself (and I would liek to add "and my own tracker", but will abstain); these guys [2] make a good argument of how and why they do it. I personally think their presence in the article is embarrassing, to say the least, and am about to remove them. If anybody sees it fit to reinstate them, feel free. complainer (talk) 10:24, 20 August 2013 (UTC)

Reverted. Netapplications is *the* most cited source in reliable sources (3rd party) and thus *the* most notable source. Remember WP:VERIFY. Useerup (talk) 06:59, 2 September 2013 (UTC)
You realize your statement (and related edit summary) is wildly unsourced, don't you? I know wikipedia is based on authority and not on truth (I do not mean this to be polemic, it is actually what the pillars say) but you do need more than a couple of asterisks to make your point. complainer (talk) 07:31, 2 September 2013 (UTC)
I do not need to source statements on the talk page. If you want to raise the bar for quoting stat collectors and require sourcing for their notability, please go ahead. You can start by deleting Wikimedia as it is not sourced *anywhere* as a RS on browser usage share. But for your convenience here is a few examples on how reputable media chose to quote netapplications: [3] [4] [5] [6] [7] [8] [9] [10] [11] [12]
No, but you do need to source what you write in the main article; i.e., if you decide to include, and put on the same level as the others, a counter which is notoriously biased to the point of being fictional, you should use the standards of care described in WP:fringe; even the CNN article you quote here, in spite of probably not having been written by some kind of web geek, expresses perplexities at the results. complainer (talk) 17:24, 2 September 2013 (UTC)
Notoriously biased to the point of being fictional? Citation needed for that claim, please. I repeat: What matters is which source is most often used by reliable sources with meaningful editorial oversight. For what it is worth, net marketshare is the only source that considers the bias of their own sample data and tries to compensate. You may disagree on their methodology, but numerous reputable reliable sources do not. That's what matters. Useerup (talk) 17:41, 2 September 2013 (UTC)
1 - you said yourself there is no citation needed in talk pages. 2 - there is a citation up there. 3 - being cited (or mentioned) in reputable sources does not make you a reputable source: plenty of reputable sources quote the Genesis, which does not mean we should support creationism on wikipedia. As for me disagreeing on methodology, no, that's not it. I think the methodology they purport to use could somewhat work. What I think, and I am in good company, is that they are doctoring data by cherry-picking sites such as Even considering all their normalizations and assumptions, there is no way in Hell that objective stats for their site can show MSIE being at twice to thrice (and Chrome being at half) the level of every other stat aggregator on Earth. complainer (talk) 18:20, 2 September 2013 (UTC)
1 - you still need to minimally justify your allegations. I supported the stance that net applications was definitively noteworthy by referring to the fact that net applications is the most often cited source for usage shares in with mainstream and computer news sources. You, on the other hand, have not supported your wild claims about bias and net applications being fictional in any way. 2 - what citation? be specific, please. 3 - yes it does. Being referred to as the reputable source by a number of tech and mainstream news organizations with real editorial oversight means exactly that. No amount of debate on Wikipedia can change the fact that the world outside Wikipedia clearly view net applications as a reliable source. That is what WP:RS is all about. As for your suspicion towards their data, please consider that *all* of the other sources are heavily biased in one form or another, e.g. mostly US sites, Wikimedia (raw data). Net applications is the only source which tries to offset the sampling bias (which probably is also why most journalists will gravitate towards that source). Useerup (talk) 18:50, 2 September 2013 (UTC)
0 - please indent properly: this makes it hard to tell who is saying what. 1 - there is a clear policy in wikipedia that says you have to follow common sense before everything else; albeit you ignored that part of my post, claiming three times the MSIE usage as any other counter is as far from common sense as one could possibly get. As for the citation I was talking about the fact that this thread is called "Source for faking" should be a clue; but let me quote something else: there’s no way of knowing whether the sites using Net Applications services are representative of the Web as a whole; in fact, they’re probably not, since 76 percent report they participate in pay-per-click programs and 43 percent claim to be commerce sites.; which comes from one of the articles you quoted. Now, thanking you for supporting my point of view, and that of anybody who has ever looked at an access counter, I'll proceed to delete the Net Application fairy tales again. complainer (talk) 07:28, 3 September 2013 (UTC)
1- Your edit simply to quash a statistic from a source which you don't like because you believe it to be biased. As I have documented, that is not a widely held position. The notability of the source is verifiable and I have provided you with a long list of references: Several reliable sources quote net applications as RS. You have offered no argument to not quote net applications beyond your personal beliefs. 2 - Furthermore, your edit leave the page in an inconsistent state. The section (Differences in measurement) just above summary section where you are edit warring contains a reference to net applications which is now an orphan. Incidentally, that section explains the different biases, which is the proper way to handle potential bias as all the sources have bias. 3 - Please restore the longstanding section until this issue has been resolved. You are engaging into edit warring. 4 - Why did you write "If anybody sees it fit to reinstate them, feel free" when you enter straight into an edit war when someone actually does just that? Useerup (talk) 5 - We now have situation where the front pages of zdnet, computerworld and several other reputable tech medias report on statistics from a source that you have decided is too biased. That situation is untenable. I will mark the section unbalanced Useerup (talk) 12:51, 3 September 2013 (UTC)
I don't think me deleting those fantastic stats after four days and four posts on the talk page really qualifies as an edit war. The revert that occurred between your two edits was not mine (I never edit anonymously)--a clear sign that I am not the only one to believe it is biased. What I have been trying to say is, actually, that the majority of the sources you provided express doubts at the objectivity of the statistics. On the other hand, if you search "Net Applications" reliability (mind the brackets and the exclusion, or you'll be swamped by MS pages), you'll find that almost no web consultant out expresses such doubt: they all think they are garbage. As for reinstatement, I suppose I expected to hear something I wasn't aware of, not to be repeated what Net Applications and Microsoft say in their press releases: I had read those already. complainer (talk) 13:04, 3 September 2013 (UTC)
Great. Then you won't mind that I reverted it. Please let it stand that way until this dispute has been resolved. Useerup (talk) 13:14, 3 September 2013 (UTC)
Aside from the fact that nothing whatsoever in what I wrote indicates that I don't mind, did you actually apply for conflict resolution before reverting the same edit twice in 24 hours (which, yes, borders on edit warring)? I certainly didn't receive any notification as a part involved. As for letting it stand, neither of us and, by te look of it, none of the other editors actually believes Net Application provides accurate numbers, as far as I can see: the issue is whether their particular brand of fairy tale is believed or not by the world outside wikipedia, and whether this justifies its inclusion. While I realize this is how wikipedia works, the conservative approach is leaving misleading data out of the main space. complainer (talk) 13:46, 3 September 2013 (UTC)
Question: How can you defend not show net applications when net applications is the source reported frequently in tech media? The latest on <a href="">computerworld's front page today</a>. I am not interested in a discussion about the sampling bias of net applications (all of the sources have sampling bias and if we were to exclude sources with sampling bias they would all have to go). I want to know why you will disregard WP:VERIFY and WP:RS. Disclosure: I would like to know this answer because this will be the central issue when I bring it to Wikipedia:Neutral point of view/Noticeboard Useerup (talk) 13:53, 3 September 2013 (UTC)
Answer: I do not defend not showing them. I defend separating a - stats by visit and stats (allegedly) by visitor; and b - fact from fiction. As a cultural phenomenon which is, as you pointed out, mentioned in many places, I do think the Net Application stats belong somewhere on wikipedia, possibly in this very same article. Where they do not belong is a table together with four other sources they directly contradict beyond any possible normalization or statistical variation. You keep throwing articles at me, but I have to ask a question myself: do you read them in their entirety? The latest one ends: Net Applications measures browser usage on smartphones, tablets and personal computers by tabulating approximately 160 million unique visitors each month who browse to [sic] the sites it monitors for customers.; even if one sees through the shaky grammar, this is really not a brilliant endorsement, now, is it? Coming to the current state of the article "Do not remove stats until the reliability of all sources has been established" is mock wikipedese. An actual reference to actual wikipedia policies would sound like "Do not insert dubious statements until the reliability of all the sources has been established"; there is no such thing as a source for removal of material, and we have no duty of granting fringe material the benefit of doubt. Incidentally, you are getting the wrong noticeboard. What you need is Wikipedia:Dispute_resolution. Bringing it to the obscure Wikipedia:Neutral point of view/Noticeboard means you are unilaterally assuming I am biased and you are not. In a one-to-one discussion, you simply have no right to do that; you also have no right to ask me why I am disregarding WP:VERIFY and WP:RS, because I am not: those policies impose that inserted material be verified, not that verified material be inserted. The distinction might be subtle but, just as with the Computerworld article, is clear to people who read them down to the bottom; and, having now accused me of at least four things I didn't in the least do, you might want to also read wikipedia: Wikilawyering. complainer (talk) 19:02, 3 September 2013 (UTC)
Some points:
  1. If the distinction unique users/page impressions is your concern then add that point to the article rather than directly deleting the stats. You obviously didn't even read this article, because the section just above the summary table where you deleted the Net Applications row makes that very point!
  2. Net Applications belong in the table. It is trivial to demonstrate that Net Applications is *the* most often cited statistics by reputable sources with meaningful editorial oversight. Debate about how much or why one or more of the statistics varies from the others does not change the basic fact that month after month tech media use Net Applications as basis for their reporting.
  3. If there are points to be made about potential bias of one or more of the sources, then you should have no problem finding sources for that and work the concerns into the article, with references. I will note that ALL of the sources have selection bias. Statcounter is biased towards European usage, Wikimedia is bordering self-referencing, Statcounter, clicky and Net Applications measure primarily hits on commercial sites. The only provider which attempts to compensate for bias (country bias) is Net Applications. All of the others report with whatever bias they have. So bias cannot be a reason to remove a statistics counter.
  4. No, NPOV noticeboard is the correct board. We obviously disagree on the bias of these sources. Suppressing a verifiably reliable source because you don't believe it can be correct is very much a NPOV issue to me.
  5. WP:VERIFY states "When reliable sources disagree, present what the various sources say, give each side its due weight, and maintain a neutral point of view.". I am not sure that there is even a disagreement between the sources (stats providers) because they do not claim to measure the same thing. But even so, it is paramount that the sources are presented. Feel free to point out the different ways the sources measure usage and how (if) they deal with potential sampling bias and what they claim to demonstrate.
  6. WP:NPOV states in the very first paragraph: "Editing from a neutral point of view (NPOV) means representing fairly, proportionately, and, as far as possible, without bias, all of the significant views that have been published by reliable sources on a topic". In my opinion, suppressing the most often cites statistics is not representing fairly and proportionally all of the significant views. Net Applications stats are regularly cited by independent reliable sources (indeed I claim that it is the most often cited stats but that's besides the point). You want to suppress the stats. I say that would be a gross violation of NPOV. If we cannot agree I believe we should solicit opinion from the noticeboard that deals with NPOV issues. Useerup (talk) 20:29, 3 September 2013 (UTC)
Useerup (talk) 20:29, 3 September 2013 (UTC)
While Net Application is seems to be "notable enough" for being quoted, I think that you're misrepresenting the bias issues. Statcounter is in fact stronger biased towards US usage than European. Their sample sizes per country, compared to the population, reveals low peneration for almost all of Asia, Africa and parts of Latin America. Low Internet penetration for large groups of people explains a lot here. Almost all European countries (except for Serbia and Iceland) have lower Statcounter penetration than the US and Canada. It is also misleading to assume a strong commercial bias; lots and lots of small noncommercial sites use the free version of Statcounter, giving the service a very broad scope. OTOH, Net Applications is very strongly biased towards commercial sites and media. They do try to adjust for country differences, but that fails for China where there is a very large reported "Internet population", which is likely a different thing that really active users of the Internet. Another issue is that a major part of Chinese Internet usage is mobile, which makes it very misleading to multiply low desktop usage statistics by a "Internet population" factor. I'd say that both Statcounter and Net Applications have very scarce data for China, making both unreliable sources for stats for this big country. At least Statcounter does not pretend that they can get to a representative number by applying a large multiplication factor. Net Applications data would be much improved by reporting China separately. (talk) 13:27, 4 September 2013 (UTC)
I just read another ComputerWorld article. It looks like Net Applications is a notable source and the article explains the different methods used. When reliable sources disagree, I believe the practice encouraged here is to report both, summarize that they vary. So netapp which reports unique users only reports that... and other which reports it reports visit counts as well. Cantaloupe2 (talk) 22:39, 3 September 2013 (UTC)
I don't think anybody doubts that Net Applications is notable; as far as I am aware, their page is nowhere near an AfD review and I myself think it is about as notable as Harry Potter, in nature if not in how many people talk about it. The issue here is whether it is reliable. complainer (talk) 16:42, 4 September 2013 (UTC)
I think it's been established in this thread that Netapplications deserves mention in the table, since reliable sources use it. The article in fact goes to some length to explain that the measurement of usage share can be difficult so we shouldn't even expect all the results to be similar. Any differences in methodology can be handled simply by disclosing them to the reader. Cheers, --Dailycare (talk) 19:37, 5 September 2013 (UTC)
The main reason for NetApplications' divergent statistics appears to be China. China has the largest internet population, more than double the size of America's, and, as other traffic analyzers show, IE is the most common browser in China.[3][4] According to StatCounter, IE has a 48.6 percent market share in China. But the traffic analyzers we use don't record a lot of Chinese web traffic. NetApplications seeks to make up for this by weighting data from each country by that country's internet population. While NetApplications' global IE market share of 57.6 percent is still significantly higher than what StatCounter records for China alone, and while there isn't another major country with a higher IE market share than China (IE is still #1 in the US, but has only a 40.5 percent market share, according to StatCounter), there is logic to the claim that weighting by national internet population would result in an increase in IE's market share. If you can cite a reliable source (not an opinion blog post) that provides evidence for the numbers being wrong, feel free to cite it. But that blog post was actually a case for Linux having more than 1 percent market share, and even StatCounter lumps Linux's narket share into an "other" category somewhere below Android's 1.72 percent. Pdxuser (talk) 23:40, 21 September 2013 (UTC)
Or, more specifically, the main reason is that NetApplications pretends to believe the CIA factbook, which pretends to believe the Chinese government's statistics that claim 43,2% penetration of internet in a country where this is a good estimate of people with access to electricity. No matter: now we have wikipedia pretending to believe Computer World pretending to believe Microsoft pretending to believe NetApplications pretending to believe the CIA factbook pretending to believe the Chinese government. I'll pretend to assume good faith and move on.complainer (talk) 12:28, 1 October 2013 (UTC)
I've just looked at this page and it really is a mess, particularly the Historical usage share section. A load of different counters saying entirely different things all shouting for attention. NetApplications clearly has a better methodology in two ways: 1. it counts visitors not hits, which is in line with the lead which talks about "percentage of visitors". 2. it weights by country. But I'm not saying that it's reliable or the one to use. What is very obvious is that the world is not homogenous in any way on this issue. e.g. in N America, IE leads Chrome by 35/31, in S America, Chrome leads IE by 61/18 and in Europe Chrome leads Firefox by 37/25. So global figures are virtually meaningless. (I'm using for these. Asia's tricky because its stats don't add up: it says Chrome beats IE 48/25 but in China it's 29/49.) Given this, what's the point of a large number of month by month tables with percentages to 2 decimal places? They aren't giving any meaningful information about which browsers are winning and the information would be far better given graphically. Chris55 (talk) 23:27, 21 November 2013 (UTC)
Winning what? Do they get a prize? There is value in giving decimal places so you can compare one month's values to the next, which shows whether the browser is gaining share or losing share. But I do agree that we only need 1 decimal point to show this, and there would be less information overload if the values were given quarterly rather than monthly. -- Schapel (talk) 23:53, 21 November 2013 (UTC)
Well people certainly aren't interested in the losers. If they aren't interested in the winners we might as well delete the page. Having looked a little more carefully, it's obvious that the situation is changing extremely rapidly worldwide and many of these statistics are out of date, particularly those for Net Applications. In the last year something like 400 million new users have come online in China and most of them use Chrome not IE, so the IE share has dropped from 74.3% to 45.9% according to StatCounter. This drowns most of the other statistics, which is why I suggest that we need to totally reorganize these tables to emphasize the regional totals. Chris55 (talk) 17:06, 22 November 2013 (UTC)
It isn't a race. There aren't winners and losers. There are browsers. I don't think anything is happening now that hasn't been happening for quite some time. IE's share is dropping because people don't like that browser, and they do like Chrome. This sort of shift in browser usage has been happening ever since browsers first appeared. -- Schapel (talk) 14:15, 24 November 2013 (UTC)

Interesting that Net Applications changed their reporting and backdated to just before Google Chrome was released. ...and every since it has maintained IE as the most used browser. Wonder where they receive their most funding Interesting that isn't it. Frankly, I'm on the side of leaving it in. Users are not stupid. (talk) 16:29, 30 December 2013 (UTC)

I'm not particularly sure that's the issue. They document a change in how they weight countries in 2007 that includes relative usage by population which means countries like China are going to rate higher in usage statistics than they do on other direct reports (especially sites rating on stats potentially skewed by the Great Firewall). On top of that, by their own FAQ, a not insignificant portion of their statistics gathering is done on Enterprise sites( Corporate, gov, orgs, medical, etc.). In particular both of those things would heavily favor Internet Explorer and Windows XP because of the legacy. All in all that means it likely more accurately reflects what is known reality (I won't bother to document this, google, it's extremely well established) but, does not, for any value of usefulness of this information, represent effective market reality for the purposes of support or general personal usage. --Karekwords?! 09:24, 1 January 2014 (UTC)

Removing the largest available and widely cited visitor stats provider on an unsubstantiated whim is ludicrous and shouldn't have been done. Thanks to those that reverted. At best it was unwise; at worst it smacks of an anti-IE bent. I see a rep from NetApplications has just kindly offered to answer any queries about their methodology below - see "Regarding Weighting and Net Applications". Psdie (talk) 18:54, 1 February 2014 (UTC)

I'm assuming you mean me since I think I'm the only person to have done anything in that section, and the one who added it. I'm in no way associated with this company, I'm a web developer and frequently have to deal with the state of web browser usage in regards to client offerings. I added the section because there was a mild revision war going on before I edited the section I reference. --Karekwords?! 03:48, 12 February 2014 (UTC)

daily browser market share stats "prove" that "Internet Explorer Users Are Only Using It Because They're Forced To At Work"

[5] Interesting. --Atlasowa (talk) 18:47, 16 February 2014 (UTC)

Wikimedia number do not add up (to 100%)

Do not seem to add up, here, under: "Wikimedia (April 2009 to present)".

Was just updating: [6] and [7]. Maybe kind of pointless, Wikimedia, not representative info there?

Minor point: In underlying data: [8]

Second: "Browser versions, mobile"
Mozilla 5.0 1,338 M 0.59% 104 M 0.50%
Mozilla 5.0 1,296 M 0.57% 113 M 0.54%

unexplained why in two lines (and higher up in first: "Browser versions, mobile", percentage add up but not "All requests" for millions). comp.arch (talk) 21:37, 5 March 2014 (UTC)

Besides not adding up, I see now that here, Mozilla is added under Firefox. That seems to be wrong (see my second edit above, where I moved Mozilla under other, revert if wrong). All Firefox versions since 1.0 are summed up under Firefox, so Mozilla is not a default user agent of Firefox. All Chrome versions since 3.0 and MSIE since 5.01 also show up in THIS year. I wander then, what Mozilla means here. It should not be the default user agent of SeaMonkey (not sure about Camino or other Firefox-like browsers). Would people be changing the user agent string manually (some extension)? As likely to happen with non-Mozilla browsers? Note that Firefox is only 11.53% and Mozilla (5.0 mostly, 4.0 and 4.5) is a lot then at 7.21%, unexplained. comp.arch (talk) 22:11, 5 March 2014 (UTC)

Wikimedia stats are tricky to deal with, the data is very raw. The Mozilla problem seems to be quite recent in Aug 2012 it was 0.46% [9], a year later it had grown to 3.99%[10] and now it is 7.21%[11]. As Mozilla is a catch all for all unrecognised browsers this looks like its not recognising the some other browser. My guess is its IE 11. Looking at my personal access logs I'm seeing
Mozilla/5.0 (Windows NT 6.1; Trident/7.0; rv:11.0) like Gecko
which has changed from the format used for previous versions
Mozilla/5.0 (compatible; MSIE 9.0; Windows NT 6.1; Trident/5.0)
I've raised bug 62311 so see if the devs can look into this.--Salix alba (talk): 10:17, 6 March 2014 (UTC)
Thanks, I overlooked that MSIE 11 was missing (should have known, they made a User Agent change to "like Gecko"). Hopefully the appropriate string will be identified and changed to MSIE 11 as all of Mozilla doesn't probably belong there. See my edit [12]. Here subtracting about 7% from Firefox gives a very low number. Then again IE is very low and probafly more right if added there. Note the all numbers seem a little low because of high Mobile numbers (not included in the page I'm editing). comp.arch (talk) 18:10, 6 March 2014 (UTC)

Spam box!

Non-mobile web browser statistics on Wikimedia
Chrome (Blink)
Internet Explorer (Trident)
Firefox (Gecko)
Safari (WebKit)
Opera (Blink)
Non-mobile web browser usage for Wikimedia visitors as of February 2014.[1]

I would like to revert this edit, as a bargraph of recent Wikimedia stats seems perfectly reasonable to me as one of the many illustrations in this article. However, given that people 'adding or restoring' information have to take responsibility for it, I find that I am at a loss to see how the figures in the bargraph have been derived from those in the cited source. I wish people would stop massaging figures in this page by applying their own secret 'corrections' and 'improvements' to them. It makes it impossible for other editors either to help maintain the page, or to see when some figures are just wrong due to vandalism or honest editor error. --Nigelj (talk) 12:05, 17 April 2014 (UTC)

It was I who updated this table to the current form. Note section above about Internet Explorer 11. I don't really like massaging the data, I wish the base data was corrected. Hopefully it will be soon. Then it's another story if this is a graph we want there. See also Wikimedia section below. comp.arch (talk) 15:51, 18 April 2014 (UTC)

The world map paints the wrong picture

The world map (the first image in the article) paints a completely wrong picture. Much too black and white, or rather much too green, blue and orange. Due to the solid colours, it looks like the usage share of the other browsers in a given country is marginal.

It would be better to assign the primary colours to the biggest three and blend them, desaturating to account for other browsers. I realise the result will have less Mondriaan appeal, but it would be much better in terms of neutrality and accuracy.

If that isn't possible, it would be better to remove the image entirely. — Preceding unsigned comment added by (talk) 22:54, 14 May 2014 (UTC)


Internet Explorer has always been the most popular browser in China. So the map is wrong with Google Chrome and I doubt that those few people in North Korea who have access to the Internet use Google Chrome. -- (talk) 03:30, 3 September 2014 (UTC)

Looks like IEs hold on China is about to be ix nixed   Doesn't look good for IE if the government is banning new versions of windows for the government.   Daniel.Cardenas (talk) 20:50, 22 May 2014 (UTC)

It's only the government because Windows 8 seem to store more preferences and this can't work for top secret government stuff. The rest of China doesn't use special browsers. They don't know about Internet Explorer's security gaps and just stick to the default Windows browser. -- (talk) 03:36, 3 September 2014 (UTC)

Opera lover

The graph that shows the proportion of browser share over time (hosted on commons with name "Usage_share_of_web_browsers_(Source_StatCounter).svg" is flagrantly biased towards Opera. In the last few years Opera share has fallen from ~3% to ~1.4%, but because of the outrageous decision to offset "zero" from the x-axis on the graph, Opera share is contrived to look stable.

Are we expecting negative browser share in the future? If not, the graph should be fixed. — Preceding unsigned comment added by (talk) 14:55, 4 September 2014 (UTC)

True, this is (perhaps unintentionally) misleading. The axes can be properly rendered by adding an extra argument yaxs="i" to the plot command (without changing anything else). See the File-page of the graph for the rest of the code.
plot(browser.years,, type="l", ylim=c(0, 70), yaxs="i",
— Preceding unsigned comment added by (talk) 15:45, 21 October 2014 (UTC)
I agree. The X axis of the graph should be at 0 and not at -3 or wherever it is right now. It looks misleading. --Maxl (talk) 20:09, 21 December 2014 (UTC)

Error for Bangladesh

Most users in Bangladesh uses Chrome, not Opera. Plz fix this. (talk) 22:04, 7 January 2015 (UTC)

Page error: white on green background does not display on mobile

Green background doesn't work in mobile view, so you're left with white on white invisible cell contents. All of the green cells appear blank on mobile, as if you have no data from StatCounter about Chrome ever, for example. I knew that was impossible. Not everyone would. — Preceding unsigned comment added by (talk) 08:26, 13 January 2015 (UTC)

Graph comparison to Unique vs Page hits

I think the graph illustrates the difference in measuring well between unique vs not unique page hits. An unregistered user removed it, I'm not sure why. If someone can explain why it should only show page hits in terms of measuring popularity of browsers I would like to see the reasoning. Personally I think unique hits is the most illustrative but I think having both is the most logical. — Preceding unsigned comment added by ScottDNelson (talkcontribs) 23:08, 18 March 2015 (UTC)

Your graph is already there, look at your graph. However, the graph is misleading when it is located at the beginning of the article. --Frmorrison (talk) 00:02, 19 March 2015 (UTC)

Why is it misleading at the beginning? If your argument is that it shows one kind of measurement (unique views) then I believe the other graph should be taken down as well since it's only showing one kind (page hits). Please clarify your point because I'm not seeing the logic. — Preceding unsigned comment added by ScottDNelson (talkcontribs) 18:06, 19 March 2015 (UTC)

Net Applications uses unique views and has widely different information than three other major firms that use hits. While Net app's data is important, it is a minority. The hits is the majority view, so it is only shown at the beginning. --Frmorrison (talk) 20:58, 19 March 2015 (UTC)

PC vs. desktop browser

I see a caption was changed to "Usage share of PC browsers" (inconsistently). People may view PC and desktop as synonymous. Note, I view "desktop browser" as same a "laptop browser" as the software is the same (or desktop OS for laptop OS (there are exceptions..). But, I view PCs as covering tablets and "mobiles" (smartphones). And "consoles" are also in a way "PCs". I could argue that we want to count all (and just say "Usage share of browsers"). StatCounter at least allows this. NetApplications link counts desktop. When counting more than just desktop[13] Chrome drops however to just 40.97%, but not Android is at 7.08% and could fairly(?) be added to it. [Similar numbers if Consoles are also added. Kind of unfair to Chrome..] comp.arch (talk) 15:59, 23 March 2015 (UTC)

I changed PC browser to desktop browser, because "PC" is inconsistent. The issue is some of the sources do not freely allow desktop/mobile to be combined, therefore to compare the stats fairly only desktop is shown.

Statcounter is free to combine the stats and Wikimedia has three sets of stats to combine. I would have to combine Chrome and Android stats, because they have the same engine and combine Safari with iPhone and iPad since most people use Safari on those devices (however, they can use Chrome and in a few months Firefox, but it would still show up as an iPad/iPhone using those browsers). Mozilla goes with Firefox, no question. I will combine those two stats to see how it looks.--Frmorrison (talk) 22:32, 23 March 2015 (UTC)

Mobile browsing on WP's website more popular as of January(?) getting more popular – something to track closely

"Non mobile" on Wikipedia:

2014-01 64.0%
2014-11 62.3%
2014-12 59.8%
2015-01 48.8% 59.3%[14] Note: numbers seems to have been updated

Mobile ("Mobile applications" even excluded): 7.68% (tablets) + 42.4% ("other mobile") = 50.08%

Yes, this is a tiny majority.. should I wait with putting this in the article? Any reason to think the numbers are wrong/they changed methodology? Or it might be a fluke..? Or am I counting the wrong numbers?:

This is for "All requests", the much fewer "Html pages" still show majority 59.9% for "non mobile".

[I was just looking for the February numbers (not in) and this doesn't quite generalize just yet.. Know other major examples/sites w/mobile majority? Still, interesting.]

See also a section I made: Usage_share_of_operating_systems#Crossover_to_mobile_getting_more_popular (note, web use is used as a proxy for OS popularity).

Of all operating systems (but I excluded consoles, hardly changes anything, I checked):

Mobile 31.58% (excluding tablets, I tend to think of mobile with..) + tablets 6.17% = 37.75%, desktop 62.26%[15]

[Update: The numbers above are higher than the combined mobile+tablets below, I had forgotten they have platform comparison.]

Conservative mobile: 21.39% (Android) + 11.24% (iOS) = 32.63%[16]

Some (much?) of Win8.1 (9.66%) and Win8 (and Win7? 33.41%) should count as mobile/tablets? But not so much conversely for the 32.63%, I acknowledge that not many use Android as a desktop, while it is possible.. comp.arch (talk) 10:20, 10 April 2015 (UTC)