Talk:Usage share of web browsers/Archive 3

Archive 1

Archive 2

Archive 3

Archive 4

Archive 5

→

Archive 7

wikimedia 94%

Why do the wikimedia stats add up to 94?

46.45% + 29.33% + 9.00% + 5.49% + 3.19% + 0.54% = 94%

Should there be a column for 6% unknown? Seems significant. Daniel.Cardenas (talk) 12:07, 19 October 2010 (UTC)

Looking at the original data, it appears to be because 4.5% of the traffic was from mobile browsers. Should we normalise the Wikimedia non-mobile data to 100%? How do the other data deal with the mobile/non-mobile split? --Nigelj (talk) 13:00, 19 October 2010 (UTC)

I suggest add a column for mobile browsers. Analysts expect this number to rise sharply. Eventually though it won't matter since the mobile browser will be equivalent to the desktop cousin. Thanks so much for adding the wikimedia data! :) Daniel.Cardenas (talk) 13:40, 19 October 2010 (UTC)

I do also think that we should add an additional column with the "lost" data and create a note where it is explained that the other companies likely do it the same but the mobile browsers don't have enough percentage of the market. (see opera mini: it is already in many stats although it is a mobile browser. some companies give extra stats for mobile browsers!) mabdul 13:46, 19 October 2010 (UTC)

Remove Statowl from summary

This was discussed earlier, see above. There is debate about whether statowl should be removed from the summary table since it doesn't represent a world view. Could you please state your preference for removing or keeping non global data from summary column. If we found a reliable source that was primarily Europe data, would you want that added to the summary table? Thanks, Daniel.Cardenas (talk) 13:54, 19 October 2010 (UTC)

• Remove from summary. I think on balance that we have plenty of sources of worldwide data. Adding one from Europe for 'balance' wouldn't work, as there's much more to the web than the US + Europe. --Nigelj (talk) 16:36, 19 October 2010 (UTC)

  I've just had a look and it really is an outlier. It gives the highest figures we have for IE and Safari and the lowest ones we have for Firefox, Chrome and Opera. No other stat provider hits all the extremes like this, some may be the highest or lowest for one browser, but have unremarkable figures for all the others. --Nigelj (talk) 21:45, 19 October 2010 (UTC)

• Keep. As an encyclopaedia the pages are suppose to be a source of broad ranging information. If the only objection to statowl is that it's US centric then I don't see the problem. Inform the "reader" of the bias in the source and let the reader decide for them-self what to do with the information. What if the reader was coming to the page to find a US biased source? When editing any wikipedia page you must always think of the reader. Henceforth, pages should be written in such a way to give the reader as much reliable information as possible. Now if statowl is being deceptive or un-scientific in their statistical reporting then that would be a reason for removal, but just selecting to report on the US -- no. Furthermore, the addition of statowl is not effecting the median. Jdm64 (talk) 18:51, 19 October 2010 (UTC)

Yes we are thinking of the reader. We are suppose to present a world view. Statowl is not a world view. We should keep the details but remove it from the summary table since it doesn't represent a world view. Would you be in agreement to add a European centric source to the summary? Why do you believe it is not effecting the median? Daniel.Cardenas (talk) 19:35, 19 October 2010 (UTC)

I'm not suggesting removing all mention of Statowl from the article, only from the part of the summary in which we determine means and medians, as its bias cannot be allowed for in those calculations. --Nigelj (talk) 19:43, 19 October 2010 (UTC)

Nowhere in the article does it explicitly state that sources must have a world bias only and nowhere does it state that a more US based source can't be included. Someone reading this page comes here for the purpose of finding what different sources are reporting. Are we going to create another page for each major country? "Usage share for US", "Usage share for...", etc.,.? That would be silly. Why not include all the sources we can find? Furthermore, if you sort the table it is clearly obvious that each source is bias. The problem is that we just don't know what those biases are, so removing StatOwl for being bias is being inconsistent because each source is biased. And I did the calculations for what the median would change to with removing StatOwl and the change is less than 0.8% for all browsers except for chrome that had a change of 1.2%. Jdm64 (talk) 01:12, 20 October 2010 (UTC)

I'd be ok with adding back both AT Internet Institute and Webmasterpro which are European biased. They're just sources. And the more we have the better the law of averages will work like Schapel said. It's better to give the reader more information which they can ignore, than to give less information so they have to continue to search. Jdm64 (talk) 01:12, 20 October 2010 (UTC)

See below comment under Schapel why law of averages doesn't apply and wp:npov requiring bias to be removed. Daniel.Cardenas (talk) 16:37, 20 October 2010 (UTC)

And if the law of averages doesn't apply, then I don't think the listed averages will be that accurate anyways, because it's not just StatOwl that is biased. Stat Counter shows unproportionately high values for Firefox/Chrome and low values for Safari. But I guess there's one benefit to removing StatOwl, then the mean/median will show the bias that I want which is higher Firefox/Chrome/Opera and lower IE/Safari values. Jdm64 (talk) 18:31, 20 October 2010 (UTC)

We are trying to remove regional bias. From that perspective do you agree the stats have less regional bias removing Statowl? Daniel.Cardenas (talk) 22:25, 20 October 2010 (UTC)

• Remove. As others have said, Statowl is primarily US-based while the other sources in the summary table are global. Keeping it messes up the averages. I also don't think Stat Counter needs a note explicitly saying that it's global, if it's implied for everything else.
  --Gyrobo (talk) 21:15, 19 October 2010 (UTC)

• Keep. As biased as StatOwl is, others are biased in the other direction. The law of averages says that with more sources the mean of the samples will converge on the true mean. -- Schapel (talk) 22:07, 19 October 2010 (UTC)

The law of averages doesn't apply. In order for it to apply the bias has to be random, which it isn't. The mean will converge towards the most bias. WP:NPOV says we must present the data with out bias. Daniel.Cardenas (talk) 03:40, 20 October 2010 (UTC)

• Remove. This page (or at least the summary table) is for global figures, not regional figures. If Statowl is kept, then the bias would need to be balanced out with regional figures from other parts of the world - Europe, China, South America, India, Russia, Africa, ..., which would obviously be crazy. Rwxrwxrwx (talk) 08:24, 20 October 2010 (UTC)

• Keep. It is a valid and reliable source, and there is no reason to discriminate against it. Wikiolap (talk) 19:27, 20 October 2010 (UTC)

  You may notice above that the reason being proposed to remove it is that it's only a reliable source for US usage, and Wikipedia needs to take a global view. Have you any reason to keep it in the light of that? --Nigelj (talk) 20:53, 20 October 2010 (UTC)

  Please review the discussion. The proposal is to remove it from the summary table because need to present a world view and it doesn't represent a world view. Daniel.Cardenas (talk) 22:17, 20 October 2010 (UTC)

I have reviewed the discussion and the reasoning about US centric vs. world view, but still feel that StatOwl should stay in the summary table. As Schapel stated - every source is not perfect in some sense - some are more country centric, others have bias toward certain web site types. StatOwl is a reliable source and should be kept, we even have a footnote saying that it is US centric - so full disclosure here.Wikiolap (talk) 15:36, 21 October 2010 (UTC)

• Remove. This page (or at least the summary table) is for global figures, not regional figures. Please remove the local statistics like statowl in the summary table. Sandro kensan (talk) 17:56, 22 October 2010 (UTC)

Summary of the votes: 5 removes and 3 keeps. Just an FYI, don't know that it means anything. Daniel.Cardenas (talk) 15:57, 24 October 2010 (UTC)

• Remove. I'll add my late vote just to make it clear. So 6 removes and 3 keeps. Daniel.Cardenas (talk) 17:12, 1 March 2011 (UTC)

• I Don't Count. Since I am the only developer working on statowl.com - I am biased - and according to the consensus here, so is my data. I just wanted to offer insight on three issues here as it relates to your debate: 1) 80% of the tracked sites are mostly US visitors, the rest are mostly Europe and South America - so we are not all US data, 2) it is so hard to get unbiased data when it comes to properly representing all geographic regions (for a multitude of reasons) - I really have tried and continue to do so despite limited means, 3) I should have a lot more Spanish content getting tracked this year - hopefully this can expand the reach further into South America/Mexico/etc. Statowl.com basically makes no money (I support the site development/hosting/finances myself) - it is a labor of love - and one that takes a lot of work. It is a little saddening to see that these efforts are not as useful as I would hope they could be to others, but I understand you guys have to do what you have to do and I am glad that it at least warranted a debate. I personally disagree with the decision to remove statowl.com data, but I am obviously biased so I won't even bother to elaborate as to my reasoning. Mikemc100 (talk) 05:16, 16 May 2011 (UTC)

Remove mean, median and the graph?

Well, some people may have won an argument to keep some 'predominantly' US statistics prominent in the article, but I fear we have all lost on the bigger purpose of providing an interesting and satisfying article in Wikipedia, with these edits. How about two medians - one for global data and one for US? (admittedly the latter would be trivial) --Nigelj (talk) 15:49, 21 October 2010 (UTC)

Yeah, I was just about to start a new topic suggesting that File:Web browser usage share.svg be given new numbers to refelct the values of Net Applications. The other graphs rely on Net Applications.
--Gyrobo (talk) 15:57, 21 October 2010 (UTC)

I suggest wikimedia be used for graphs. Netapps adjust their numbers based on CIA factbook of internet users by country. Daniel.Cardenas (talk) 16:31, 21 October 2010 (UTC)

I suggest StatCounter is the best source that reflects the current trend in usage; all the other sources seem to be converging to StatCounter. Jdm64 (talk) 19:25, 21 October 2010 (UTC)

I'm good with StatCounter if others are o.k. with that. The reason I think wikimedia is better is that we get to report mobile browser share too. Seems to be missing from the others. Daniel.Cardenas (talk) 20:24, 21 October 2010 (UTC)

We need some outside opinion about the use/inclusion of the averages/graph -- from a Bureaucrat/Steward or some one who would more likely know the exact details of WP:NPOV. I don't think that the averages/graph fall under "Synthesis". Because if you read the example of synthesis it's about skewing the sources to say something they don't to push an agenda. Instead, I think that the averages/graph fall under the safety of "Routine calculations". Jdm64 (talk) 19:25, 21 October 2010 (UTC)

Yes outside opinion would be good. I doubt there will be agreement to put them back in with a U.S. bias. Daniel.Cardenas (talk) 20:24, 21 October 2010 (UTC)

True, but we might be able to get a consensus of whether using the averages/graph is ok. Jdm64 (talk) 21:57, 21 October 2010 (UTC)

What makes a good source of browser usage

Just thought I'd jot done what thought makes for a good source of usage data. Feel free to add your thoughts.

1. Raw data. In other words not multiplied by CIA world factbook of internet users by country.
2. Reports the whole story. In other words does not exclude mobile browsers.
3. Global focus
4. Wide variety of web sites.

Based on the above criteria, all of the sources fail in one or more of the criteria. So when judging which sources to use for statistics or wp:lead figures how do editors decide which to use? There is a group here that wants to include all sources and hope that the bias will work its way out with many sources. Another group wants to eliminate sources with heavy bias. I'm not offering a solution, just pointing out what is going on.   Daniel.Cardenas (talk) 16:21, 23 October 2010 (UTC)

If two or more sources implemented all of the above criteria we could judge source accuracy by correlation. Daniel.Cardenas (talk) 19:18, 23 October 2010 (UTC)

Looking at the summary table there are two sources that don't correlate "well": StatOwl and NetApplications. And both of these have a note as to why.   Daniel.Cardenas (talk) 20:15, 23 October 2010 (UTC)

I would refrain from making subjective judgement about sources and their methodologies. As encyclopedia we need to make sure that the sources are reliable and verifiable, but not whether we like what they report or not.Wikiolap (talk) 04:41, 24 October 2010 (UTC)

That was basically my whole point about keeping StatOwl in the summary, but somehow it turned into a big argument, and now we have no mean/median and no graph using those. Jdm64 (talk) 04:58, 24 October 2010 (UTC)

A common task of editors is to decide which sources are best to include in an article. Using source with a global view versus one that isn't is a good choice for a summary that per policy needs to have a world view. The fact that the bias is disclosed doesn't help.   Daniel.Cardenas (talk) 16:03, 24 October 2010 (UTC)

Vote again - about removing regional stats in summary

Options:

1. Remove - Remove stats that predominately favor a region, because it doesn't represent a world view. Fully disclosing the bias in the summary doesn't help. We can keep regional sources in a second summary table.
2. Keep - Keep regional stats and even add more. The conglomeration of stats will smooth out regional bias.

Daniel.Cardenas (talk) 22:48, 29 October 2010 (UTC)
Before asking for external help I would like to vote again and see if opinions are the same. I'm not including wp:synth as an option because the policy states that if editors agree then it is o.k. to have simple math. Below are the two opinions that are being debated. Feel free to edit the text if you can better summarize it. If we can't agree then we need to ask for outside opinion as to which source(s) to use for intro message, because we probably won't agree on that either.

---

• Keep -- Because I see the summary and entire page as a summary of reliable usage share information, not as a summary of global usage share information only. Jdm64 (talk) 00:59, 30 October 2010 (UTC)
• Keep -- and maybe create an extra summary table. mabdul 11:00, 30 October 2010 (UTC)
• Keep -- If the source is reliable and verifiable - keep it. Every source may have some bias, whether it is geographical, certain type of sites or something else. Wikiolap (talk) 01:49, 31 October 2010 (UTC)
• Remove again. If people really want to see regional statistics, then create a separate article for each region. This article is too big as it is. See WP:WikiProject Countering systemic bias, WP:NOT#Wikipedia_is_not_an_indiscriminate_collection_of_information, and WP:Article size, Rwxrwxrwx (talk) 10:16, 31 October 2010 (UTC)
• Remove. Two reasons: (1) If we had stats covering other major regions (Asia, Africa, S. America and/or Australia), then we would have a world view. As it is we just get a US-bias in the summary. (2) The US stats are outliers in every single case - the highest figures we have for IE and Safari at the same time as the lowest ones we have for Firefox, Chrome and Opera. No other stat provider hits all the extremes like this. Of course this adds a US bias to the summary. --Nigelj (talk) 11:23, 31 October 2010 (UTC)

Both AT Internet Institute and Webmasterpro are stats from Europe. Jdm64 (talk) 22:48, 31 October 2010 (UTC)

And do they appear in the summary table? No. Please stop bringing up this strawman. No one is suggesting removing StatOwl from the page, just from the summary table. --Nigelj (talk) 09:00, 1 November 2010 (UTC)

I was responding to "If we had stats covering other major regions", which we do. Not every region, but to say that we only have access to US stats is not true. We both are using the same argument, I'm just on the other side of you. Also, the original vote mentions my argument "Keep regional stats and even add more". I don't see what you're complaining about? Jdm64 (talk) 18:23, 1 November 2010 (UTC)

• Remove. Remove the local statistics from summary table. Sandro kensan (talk) 14:50, 1 November 2010 (UTC)
• Remove Say that you want local statistics, then, how many sources do you have for each region? How big do you want the summary table? Juanjosepablos (talk) 00:47, 3 November 2010 (UTC)

Statcounter offert statistics from Continent and States, i can insert stats from USA, Italy, Germany, France, Belgium, Irland, Russia, and go on. The summary table will be enormous. Sandro kensan (talk) 21:50, 3 November 2010 (UTC)

• Remove - wp:npov is one of the 3 core content policies of wikipedia. Not presenting a world view is an unacceptable bias. Claiming all sources have a bias with out discussing the detail(s) of the bias is a cop out.   Daniel.Cardenas (talk) 16:41, 7 November 2010 (UTC)
• Keep. These are the best available sources. Add footnotes to draw readers' attention to any major bias issues.--Harumphy (talk) 08:45, 8 November 2010 (UTC)

Why is the pie chart back?

The pie chart I thought was voted out for being harder to read than a bar chart.

The 3d one is even more stupid and harder to read. —Preceding unsigned comment added by 87.74.75.109 (talk) 16:56, 1 November 2010 (UTC)

Wikipedia Squid Stats not up-to-date

Not a fault with this page as such, just one of its sources. http://stats.wikimedia.org/archive/squid_reports/ was last updated mid-October. Why no end of month data? 217.64.123.99 (talk) 10:40, 16 November 2010 (UTC)

W3counter only adds up to 90.8%

For the month of October the usage stats are 9% short of 100%. What do you think about that? Daniel.Cardenas (talk) 23:16, 11 November 2010 (UTC)

And your point? This happens regularly with other sources, maybe not at that level, but it's not uncommon. There are other browsers other than the one's we report. For example, Wikimedia is currently off by 5% and Clicky about 6%. Jdm64 (talk) 08:48, 12 November 2010 (UTC)

Again: I do think we should a new column to every chart called "others" that explains the missing percentages. mabdul 10:27, 12 November 2010 (UTC)

I agree with Mabdul. The size of these 'others' is going to continue to increase due to mobile browser usage, and should not be kept hidden. Their existence is also a good reason not to display stats in a pie-chart without an 'others' segment, the data rarely adds up to 100%, and yet is displayed as 100% of a circle, which is misleading. That becomes even worse if we try to display medians in a pie chart: even if we use the median 'others' figure, the total will not be 100%. That's why a bar-chart was better - there's no implication of 100%, of 50% not looking like 180 degrees, etc. --Nigelj (talk) 17:25, 12 November 2010 (UTC)

I'd vote for reinstating the original chart. Jdm64 (talk) 09:48, 13 November 2010 (UTC)

Better readable: dito!

mabdul 16:20, 13 November 2010 (UTC)

Wikimedia is not off by %5. Perhaps your thinking prior to the 4.7% added for mobile. Clicky is at 99.82%. Added an unknown column for W3counter. Other implies that it is "other", and we don't know that. Could be some calc error on their part that needs to be added to current list of browsers.

For next month I'm planning on using StatCounter instead of wikimedia, unless you guys suggest otherwise.   Thx, Daniel.Cardenas (talk) 04:09, 13 November 2010 (UTC)

Ok, I stand corrected, I miss read the Wikimedia source page. But the point I was trying to make was that the sources don't always add to 100%, and that depends on the methodology use for sampling. I'm just hoping you not thinking of dropping the source because of that. Jdm64 (talk) 09:48, 13 November 2010 (UTC)

As Nigelj indicates: with more users online by phone, the browser landscape will change: why not indicate this in add others? with a small explanation everybody is served... mabdul 16:20, 13 November 2010 (UTC)

I was never advocating against an "other", if that's what you thought. I was just trying to defend against removing sources, because of the recent StatOwl debate. The usage share of OSs page includes an other, and I think so should this page. But if an other were to be added, then I think it's only fair to add back the median value to the table as well. Jdm64 (talk) 20:11, 13 November 2010 (UTC)

Added the median to the world wide summary table, because there doesn't seem to be too much objection to that table. Assuming this is o.k. then I'm good with adding it back to the non world summary table also. Daniel.Cardenas (talk) 05:47, 14 November 2010 (UTC)

What's the bias of this page?

It looks like the underlining issue with the discussion above (Remove Statowl from summary) is that the purpose for the page is ambiguous and/or not agreed upon. The current wording of the page doesn't clearly state what bias is to be mandated for the page. It looks like there are two main camps. Please state what type of bias you want the page to have. (I've exaggerated the summaries to get the points across)

(1) The page is a source of world usage only, so any non-world biased source would taint the accuracy of the mean/medians of the summary table, and should henceforth be removed. The accuracy of the mean/median is valued higher than giving the reader more information and letting them decide what to do with the information. Any user that's looking for regional information that could be provided should just google for it. The key objective is to remove sources until the mean/medians are accurate. The mean is greatly effected by outliers and that is why only sources with small biases should be included.

(2) The page is a summary of usage share information from the sources (companies/sites) that gather the information. Any source can be included as long as it follows a few guidelines*. Providing the reader with information is valued higher than being overly strict and sensitive about what sources are added and how they effect the mean/median. The key objective is to provide the user with the information that's out there and summarise it in a clear and concise way. Every source has a bias, but as long as the bias is understood/stated and/or is within the guidelines* the source should be included. Because of this, the median is more useful because it's less effected by outliers.

Votes:

(2) Jdm64 (talk) 03:01, 21 October 2010 (UTC)

Per wikipedia wp:npov we must present a non biased view. The summary should contain a world view and be removed of any obvious bias. The page is not a source of world usage only. Country biased stats can be presented but should not be mixed with summary world view information. NPOV is one of wikipedia's core principles and is non-negotiable, per policy. The reader should be given plenty of info in a non biased way. Any reader looking for regional information can look in the appropriate section for it, if it exists. Daniel.Cardenas (talk) 03:41, 21 October 2010 (UTC)

I don't think you're using NPOV correctly in this case. NPOV is for things that have a View Point (for/against); statistics don't have a viewpoint. Statistics is sampling of reality; it's only the interpretation or application that might have a view point. Actually, by not including a reliable source that is trying to be added/maintained, you are actually introducing a bias and being non-neutral. It looks like we're both accusing each other of failing to uphold NPOV. Furthermore, nowhere in "Summary table" is there any mention of the fact that only world biased sources can be included. You are actually imposing a bias into the table; I'm trying to remove that bias. Jdm64 (talk) 05:59, 21 October 2010 (UTC)

Can you summarize which bias your concerned about. Want to summarize the arguments and re-vote and or request outside opinion. Daniel.Cardenas (talk) 07:00, 21 October 2010 (UTC)

Ok, here's my main arguments.

1. I view the page (including summary table) as an unbiased conglomeration of current usage share statistics from sources. While you view the page as a global only summary, as a way to create original research through the use of the summary table, especially the median/mean. (see below for why)
2. Nowhere in the page does it state that global must be upheld and that regional can not be included in the summary table. It's an point of view to think that the summary should only include global sources. Likewise, I admit that including non-global is also a point of view. But we need to decide what way we want it.
3. StatOwl is not a "hard" regional source. Their website states, "92% of web sites serve a predominantly United States market (this is an area we would looking to add more diversity in the upcoming months)". They are looking into the issue and it's not like they're excluding page hits from outside the US, but that the selection of web sites they've picked have for whatever reason been predominately accessed by US visitors.
4. All sources are biased -- it's statistics. If there was no bias then all the sources would be very close to the average. Henceforth, I feel that you're singling out one source for having a bias even-though they've done their statistical due diligent.
5. NPOV is for pages that cover non-data topics where one can form an opinion. If I say that there's X million Firefox users and Y million Chrome users, that's either true or false. But if I say that Firefox is better because it has more users than Chrome, then that's an opinion and would not be NPOV. The point being that including a, what you call, regional source can not be classified as being non-neutral because it's only a report of what a source said, not our own opinion.
6. The NPOV page states "Indicate accurately the relative prominence of opposing views", and a regional source would be an opposing view. So by not including it you violate NPOV because of the relative prominence of web sites and users from the US.
7. Our inclusion of the mean/median is in-fact original research. We don't have the authority to do statistical analysis on the sources because the data that we receive has already been "processed". Any number that we summarise with will always be an opinion. So your worry over the averages being biased (with statowl) is pointless because we're really not suppose to do it in the first place. Taking the average of an average only tells you the average of the average not a more accurate view of the real values because we can't garentee that the sampling methods used by the sources can be trivially combined.
8. Basically by including the average of only global sources we're stating that Wikipedia asserts that the current world usage share is X%. But if we look at it as we have Y number of reliable sources and this is the average then there's no assertion made by Wikipedia. The key difference is the purpose for summary. In the first case, we're summarising to conclude something new, because we feel that averaging the global sources will allow us to be more accurate. In the second case we're summarising to show the distinction among the sources (mainly the difference each source is from each other). Jdm64 (talk) 09:37, 21 October 2010 (UTC)

Ealier I felt that Statowl should have been removed because it distored the results of the mean and median. But now I feel that the inclusion of a mean/median in this article is not just original research, it's a violation of WP:SYNTHESIS, for the reasons Jdm64 mentioned.
--Gyrobo (talk) 15:16, 21 October 2010 (UTC)

(2) mabdul 16:40, 21 October 2010 (UTC)

(1) Sandro kensan (talk) 18:01, 22 October 2010 (UTC)

These votes are meaningless because the options (1) and (2) are not the two options. No one is proposing that all US data be expunged from the page, so that people looking for it "should just google for it". The question is whether to include one predominantly US dataset in the calculation of the mean and median. See Straw man. --Nigelj (talk) 20:24, 22 October 2010 (UTC)

A non-biased page would not present a graph titled "Alternate Browsers" that ignores one browser completely. Either show all or show none. What makes those "Alternate" if you treat all browser equally? —Preceding unsigned comment added by 207.238.43.12 (talk) 14:30, 17 November 2010 (UTC)

About NetApplications

NetApp measures browser usage different than the other sites. It gives greater weight to a user and less weight to a page load. I can explain further if you'd like. Because of this I suggest it be put in a separate table. One table is based on raw page loads/hits, and other is for usage share based on number of users. The advantage of this is that we aren't trying to mix apples and oranges with the statistics. What do you think?   Thanks, Daniel.Cardenas (talk) 05:43, 15 November 2010 (UTC)

do you have references for that? Does this really change anything the market share? mabdul 06:35, 15 November 2010 (UTC)

What specifically do you want a reference for? There is info on each of the web usage reporting sites. Here a link to more of an overview: http://news.softpedia.com/news/IE-vs-Chrome-vs-Firefox-Browser-Market-Share-Insight-159669.shtml   Thanks, Daniel.Cardenas (talk) 23:16, 15 November 2010 (UTC)

Net Applications is measuring world wide usage of web browsers, and as such should be listed in the world-wide table. Each source have different methodologies, and if NetApp's is different enough - we should point reader's attention to it, but not to exclude this source. It is both reliable and verifiable, and also notable (press likes to cite NetApp numbers). Wikiolap (talk) 05:26, 17 November 2010 (UTC)

I feel NetApps should be included. It may come up with slightly different numbers, but if it serves the same market as the others, and the methodology is sound, it shouldn't be excluded.
--Gyrobo (talk) 03:16, 19 November 2010 (UTC)

One more thing: now that the median value has been added back to the table, shouldn't the graph reflect that value, rather than any one source?
--Gyrobo (talk) 03:21, 19 November 2010 (UTC)

Pie Chart vs Bar

Do you guys like the bar chart because of the 100% issue, and or other reasons? Daniel.Cardenas (talk) 19:38, 23 November 2010 (UTC)

Pie charts in general are problematic. Here are couple of links to talk about it:

• http://blogs.oracle.com/experience/2010/03/countdown_of_top_10_reasons_to_never_ever_use_a_pie_chart.html
• http://www.gilliganondata.com/index.php/2009/12/02/how-succinctly-can-i-explain-why-pie-charts-are-evil/

Wikiolap (talk) 21:16, 23 November 2010 (UTC)

I agree they are problematic, but I don't agree the bar chart is necessarily better. The thing that a pie chart conveys and that a bar chart doesn't, is that all of the browsers are competing for a share of the web usage pie. Thanks for the links. Daniel.Cardenas (talk) 22:42, 27 November 2010 (UTC)

The thing we would have to show is that the browsers compete for a share of 100% of the pie. Making a collection of medians add up to 100% is not trivial. It is not acceptable for, e.g., a 50% share to occupy other than 180 deg. Tilting the chart in 3D adds further issues as thin slices near 3 and 9 o'clock then look much thinner than the same ones would near 12 or 6; with true 3D perspective, sectors near 6 o'clock can look even bigger still. These problems disappear (i.e. do not need solving) with a simple bar chart. --Nigelj (talk) 13:17, 28 November 2010 (UTC)

Currently the means add up to 100.6% so currently it is trivial. Yes there are pie chart issues, but still doesn't resolve the issue that they convey an important message: That all browsers are competing for a share of the pie and each is trying to grab the largest slice. If we took a user survey which they prefer the bar or the pie, which do you think would be preferred? Perhaps we could have both for a while. Daniel.Cardenas (talk) 16:35, 28 November 2010 (UTC)

I really find it hard to read and realizing which browser has how much market share in the comparison of other browsers in the pie chard. 3D makes it almost harder. Please: use the bar charts (and make line for at least every 20% market). mabdul 16:50, 28 November 2010 (UTC)

I don't really understand why a bar chart is used over a pie chart, it seems an illogical way to present the data and a pie chart seems much more approriate. Even if the numbers don't quite add up a proportional factor can be applied to the shares to correct the error. 193.117.31.4 (talk) 16:43, 8 December 2010 (UTC)

Did you read wikiolap's two links above? I made them more obvious. I agree, but given the preference for bar charts perhaps someday we could have both. Daniel.Cardenas (talk) 22:09, 8 December 2010 (UTC)

IE bursting out of the graph

Why does the new bar graph axis end short at 40%? I assume it's not the classic marketing tactic of making the IE bar look like it 'bursting out of the graph', rather than showing it to be 'just less than 50%', which is what it is. --Nigelj (talk) 12:18, 3 December 2010 (UTC)

Its a quirk of the default settings of statistical package used: "R". Probably can be fixed nowing the correct parameter to change. Daniel.Cardenas (talk) 22:06, 8 December 2010 (UTC)

35% Opera usage in Ukraine?

I've lived in Ukraine for 5 years and I can only think of one time I have ever seen opera used there. This made me a little suspicious of the supposed 35% usage there, so I checked the reference number and all that is referenced is "Top 5 browsers in Ukraine". It is not click-able, unlike the "Top 5 browsers in Germany" link above it, and doesn't seem to be linked to anything at all as far as I can tell. Someone needs to check up on this because I really don't think this is correct (my personal experience could certainly be wrong, but no link?). —Preceding unsigned comment added by 66.169.150.165 (talk) 03:32, 8 December 2010 (UTC)

Yep, there is something wrong with that referenced. If you look at the source you can see better what it is suppose to point at. I hacked it it is really obvious now. http://gs.statcounter.com/#browser-UA-monthly-200910-201010-bar . Daniel.Cardenas (talk) 22:03, 8 December 2010 (UTC)

The wrong table is repeated at the head of the article

Now that we have three summary tables in this article, I believe that the wrong one currently has prominence at the top of the page. With only two lines of data in it, the November table is currently inaccurate and highly misleading. The October table has eight lines of data (even though one of them is still regional) and so is much more suitable for prominence and graphing. Promoting the new month's data long before it is complete is not the best way to use the current layout IMHO. --Nigelj (talk) 22:45, 9 December 2010 (UTC)

Over and Under estimation

This section has unencyclopedic content. I'm planning on deleting most of the text that doesn't have a reference. Feel free to add it back in if you have a reference. Some of the text is just wrong. For example the comparison of text versus graphical browsers. The web hit counting software that I've seen counts page views and not image views. Daniel.Cardenas (talk) 21:06, 10 December 2010 (UTC)

I support to keep only these bullet points in this section which have references. I also want to point out, that it is not our role to decide what is "wrong" and what is "right", but only what has verifiable and reliable references. Wikiolap (talk) 19:53, 17 December 2010 (UTC)

Change date format from January 2010 to 2010-01 ?

What do you think about changing the date format from something like January 2010 to 2010-01? The advantages are:

1. Sortable
2. More compact
3. Perhaps easier to read for non native english reader.

Disadvantages are:

1. Not the typical way we write dates.

Thanks, Daniel.Cardenas (talk) 16:19, 17 December 2010 (UTC)

I prefer to keep the month names - they are more readable than numbers. And for non native English readers (I am, BTW, is non native English reader too), if they don't know month names - they probably won't know most of other words. Wikiolap (talk) 19:56, 17 December 2010 (UTC)

Wrong stats used!!

Check out the stats on

http://www.w3schools.com/browsers/browsers_stats.asp

They are not the same as in this article from November 2010!! —Preceding unsigned comment added by 86.59.94.16 (talk) 09:46, 30 December 2010 (UTC)

W3Counter or w3Schools?--Sandro kensan (talk) 20:27, 30 December 2010 (UTC)

indeed, I also totally agree that the statistics used in this article currently is totally inapropriate! Searching today for this kind of information (from studying purpose) I could agree only to the information provided on w3schools web page... I hope it'll be changed in this article soon! Or at least that source with it's data gonna be provided... otherwise it gonna look like browser lobbies reached even wikipedia :/ 78.104.123.6 (talk) 20:31, 3 January 2011 (UTC)

Can you expand on your opinion? Why only w3schools web page? Daniel.Cardenas (talk) 22:03, 3 January 2011 (UTC)

Europe FF overtaking IE in lead section pool

Does the mention of FF overtaking IE in Europe in December 2010 deserves to be in article lead section.

• Oppose - it is newsworthy right now, but there are many dynamics in browser wars at different times in different places - it doesn't fit into lead, but rather to Browser Wars article. Wikiolap (talk) 18:46, 6 January 2011 (UTC)
• Support - IE has been the worldwide dominant browser since about 1999. It's usage share graph has been on a downward trend for some time, but this is the first time it has lost its leading position since then (11 - 12 years is a long time in the history of the WWW). And in Europe too - that's an important world region. Who'd have guessed just a short time ago? This is one of the most notable milestones in web browser history. --Nigelj (talk) 19:51, 6 January 2011 (UTC)
• Support - When it is not newsworthy, it will be replaced with something similar. For example: Firefox is the most popular browser in Europe. Daniel.Cardenas (talk) 20:41, 6 January 2011 (UTC)

For the 1999 info that was reverted from the lead, that could go in a history section. Daniel.Cardenas (talk) 00:37, 7 January 2011 (UTC)

This looks like double standard - if FF overtakes IE (if only in Europe), then it is interesting. But if IE overtakes NS - then it is not interesting. I am just trying to keep article balanced and unbiased - either both deserve to be in the lead, or none. Wikiolap (talk) 01:03, 7 January 2011 (UTC)

There is a reason history isn't in the lead in most articles. Its mostly about what is relevant now. thx, Daniel.Cardenas (talk) 04:04, 7 January 2011 (UTC)

statcounter mobile numbers suspect

• http://en.wikipedia.org/wiki/Usage_share_of_web_browsers#StatCounter_Mobile

Why does RIM start so low and now overtaking iPhone? iPhone dropping in marketshare? This seems to be contrary to what everyone else is publishing. Seems like those two columns should be reversed. Wondering if statCounter has got something messed up. Any thoughts? Thx, Daniel.Cardenas (talk) 06:10, 2 December 2010 (UTC)

I sent statCounter an email. Lets see if they respond. Daniel.Cardenas (talk) 13:35, 2 December 2010 (UTC)

On Thu, Dec 2, 2010 at 10:21 AM, StatCounter Global Stats <globalstats@statcounter> wrote:
Hi Daniel,

Unfortunately, we would not be in a position to provide explanations for all the trends we see on StatCounter Global Stats - generally, we report the trends and leave the commentary and speculation to journalists, bloggers and other interested parties who make use of our stats.

In this case some other possible points that may be of interest to you include:

• Users may be moving towards using apps on the iPhone rather than the web browsing functionality
• BB appear to be improving their handsets per the latest handset reviews - making them more user-friendly for web browsing etc e.g. Blackberry Torch - "with the BlackBerry Storm touch/click screen, this touch screen is simply light years ahead and also with greatly improved web browsing functionality" (source:http://www.blackberry-phone.co.uk/)

In my own personal experience, Blackberries have been used mainly in the past for the purposes of email and not for web browsing due to the limitations of the browsing functionality. If this is changing, it would explain the stats we are seeing as you suggest.

You should also note that it's possible to use OTHER browsers on the iPhone e.g. Opera (http://gigaom.com/2010/05/28/opera-says-2-6m-iphones-owners-use-opera-mini-do-you/) so it's not necessarily the case that usage of the iPhone is dropping although usage of the iPhone browser has been falling.

Jenni

On Thu, Dec 2, 2010 at 9:54 AM, Daniel Cardenas <daniel@> wrote:
Hi Jenni,

Thanks for the reply. Here is an example of where your numbers don't make sense. In 2008 you report that iPhone had 22% browsing market share and RIM had 4.7% share. iPhone was relatively new at that time and RIM was established at that time. It makes sense that RIM would have 22% share and iPhone 4.7%. Here is a sales history of iPhone which should ruffly correlate to browser share. http://upload.wikimedia.org/wikipedia/commons/d/d9/IPhone_sales_per_quarter_simple.svg
Here is one that includes blackberry sales:
http://www.investorvillage.com/smbd.asp?mb=445&mid=5083828&pt=msg

The numbers that statCounter publish can only make sense if one assumes that blackberry's have not been used much to browse the web and now they are starting to be used to browse the web. I'll assume that is the case because that is what your numbers are telling me.

Thanks, Daniel

On Thu, Dec 2, 2010 at 5:39 AM, StatCounter Global Stats <globalstats@statcounter> wrote:
Hi Daniel,

I'm not sure what source you are using for comparison purposes with our figures but we are confident in our stats - there is no error. It's important to ensure that you are comparing like-with-like - we report on mobile internet usage for example, not on sales of handsets.

This recent press release may be of interest to you: http://gs.statcounter.com/press/blackberry-overtakes-apple-in-mobile-wars

Thanks for your feedback.

Jenni
StatCounter Global Stats
Daniel.Cardenas (talk) 20:34, 2 December 2010 (UTC)

---

and that means (for us)? what are the reasons? mabdul 20:51, 2 December 2010 (UTC)

They are standing by their data and give good reasons for it, even though the numbers may seem very odd. Daniel.Cardenas (talk) 01:23, 3 December 2010 (UTC)

I've reviewed some other websites data and it doesn't come close to what statCounter says. (No I can not supply references.) Matches closer to what wikimedia says, so I'm confident statcounter is doing something unexpected with the logs. Not much we can do about it. Daniel.Cardenas (talk) 05:36, 3 December 2010 (UTC)

In my view, this is not at all unexpected. BlackBerry always had a huge installed base of handhelds. However, they were mostly used for corporate e-mail. From my personal experience, it was next to impossible to browse the normal web with the BlackBerry handhelds; the browser was very slow (partly caused by the handheld not offering WiFi, partly because of the slowness of the processor). Only starting in the end of 2008, when the BlackBerry 8900 and 9000 were released, was the browsing experience acceptable. It only got good once BlackBerry OS 5.0 was released in autumn 2009. Given that corporation typically replace handhelds after around 2 years, this would be entirely consistent with the market shares observed. --SmilingBoy (talk) 09:59, 3 December 2010 (UTC)

Do you have a suggestion why a large family of websites such as wikimedia has mobile safari 5 times larger than blackberry, while statcounter says blackberry is higher? Thanks, Daniel.Cardenas (talk) 13:47, 3 December 2010 (UTC)

Getclicky.com also says that mobile safari is 5x blackberry. http://www.getclicky.com/marketshare/global/web-browsers/#/marketshare/global/web-browsers/mobile/   Daniel.Cardenas (talk) 01:33, 5 December 2010 (UTC)

Is in this table Opera Browser Opera Mini and Opera Mobil combined or is it separated? If not separated: why not splitting them up, if separated: why not making another column? mabdul 13:57, 3 December 2010 (UTC)

Don't know about opera mini. Seems netapp created a column for mini recently. I combined it into one column for the summary table. Don't know if mini is counted in the mobile number, since that is where mini is used, i think. Daniel.Cardenas (talk) 21:18, 4 December 2010 (UTC)

I mailed statcounter because of the opera mobile proble. here is the response...

Hi Dennis, 

We include both Mobile and Mini under Opera in our Mobile Browser Stats. 

We also include Opera usage on ANY mobile device as defined here:
http://gs.statcounter.com/faq#mobile-definition

Many thanks for your interest in StatCounter Global Stats. 

StatCounter Global Stats

-----Original Message-----
From: **@**.**
Sent: Friday, 7 January, 2011 02:58
To: globalstats@statcounter.com
Subject: Global Stats Feedback From Dennis


               Name: Dennis
               Company: n/a
               Email: **@**.**
                       
               Hello,
can you explain me what count as a Opera mobile browser? Does this include Opera
Mobile, Opera Mini and ofter devices as the Wii or the Nitendo DS?
Or is only Opera Mobile counted?

Regards
Dennis

mabdul 22:40, 7 January 2011 (UTC)

Your opinion is requested on what should or shouldn't be in the lead

wp:lead says that the lead should be a summary of the article, mention the most interesting points and be about 3 paragraphs long. Here are some things that could (doesn't mean should) be in the lead:

1. Fun facts about each browser. What makes the browser interesting in terms of usage share.
2. Regional variations in usage share. Perhaps the same as the above.
3. Latest trends in usage share.
4. History of one or more of the above.

I didn't include in the above items: definition, summary table, and chart. I'm thinking people are good with those. What do you think the lead should look like? My opinion is we should add some text about the x year trend of I.E. dropping in usage share and recent chrome, safari, and mobile gaining usage share. Thanks, Daniel.Cardenas (talk) 18:05, 7 January 2011 (UTC)

I won't add funfacts.

1. in the first summary table I would add the recent changes.
2. I would (somewhere in the page) add a summarized (svg) graph of all browsers. (like adding mosaic for the beginning...)

just my 2 c mabdul 18:41, 7 January 2011 (UTC)

• I would keep regional variations in the dedicated section. In fact we already have this section :)
• I would not keep info about specific recent changes to browser market share landscape in the lead either - if they are worthy mentioning - they can be in their dedicated section. It is OK to have some summary in the lead, like saying that IE used to be a king, but not anymore, rise of mobile browsing.

Wikiolap (talk) 19:29, 7 January 2011 (UTC)

Commentary about the mobile column

Wikimedia lists mobile for december at 6.4 percent.[1] 3.4% of this is safari. No doubt much of this is due to the ipad and itouch. I wouldn't call an ipad with wi-fi access mobile, but I believe wikimedia does. Notebooks are starting to come with 4g access, they seem very mobile, but we don't include them in the mobile column. For some mobile means using a limited browser, but even cell phone handset browsers seem full featured with flash. The category seems murky and no doubt a better way to categorize will present itself as mobile continues to grow.[2] Daniel.Cardenas (talk) 22:26, 17 January 2011 (UTC)

template

Why is the web browsers template removed? It is really a part of the article...mabdul 22:51, 17 January 2011 (UTC)

The article is about usage share of browsers, not browsers. If you think it belongs here that add it back in. Perhaps it can be minimized by default.   Daniel.Cardenas (talk) 22:56, 17 January 2011 (UTC)

This template includes many browser (not all which have an article). But there are also related articles in. and this contains the usage share of web browsers. (already linked in) mabdul 23:30, 17 January 2011 (UTC)

There are some problems about the percent in the tables

For example, Stat Counter: 46.94% + 30.76% + 14.85% + 4.79% + 2.07% = 99.41% (0.59% is other), but the mobile browsers are 4.10%.

This meant that these data (100%) are refer to desktop browsers only, are NOT contained the mobile browsers. So, these tables should add the NOTES. - 111.251.195.122 (talk) 01:45, 24 January 2011 (UTC)

I think others prefer to adjust the percentages, I was a bit reluctant to do that because for the world wide summary table it cancels out low numbers their. Daniel.Cardenas (talk) 04:26, 24 January 2011 (UTC)

NetApplication unique visitors

«Netapplications bases its usage share mainly on the usage statistics of 40,000 websites having around 160 million unique visitors.» I ask: "160 million unique visitors" is it a correct sentence?

The official netapplication site writes: «We ‘count’ unique visitors to our network sites, and only count one unique visit to each network site per day.» [3] I suppose unique visit to each network site is different from unique visit to all network sites. Or "each and "all" is the same?

So if each single site has unique visitors that is summed with other unique visitors to other network site then 160 million divided 40,000 website is the average unique visitors from one site that Netapplication monitored. 4000 is the one site average unique visitors: not much.

The 40,000 website's unique visitors is less, much less of 160 million. 160 million is a sum of unique visitors but 160 million are not unique visitors. Do you agree?--Sandro kensan (talk) 17:58, 31 January 2011 (UTC)

World Map of Most Popular Browsers by Country

The world map of most popular browsers by country is available at http://www.browserrank.com. It is updated regularly (weekly) and it is based on statcounter.com data. It also shows usage share of Internet Explorer and Chrome on country level too. It should be included as a visualization of worldwide distribution of popular browsers. —Preceding unsigned comment added by 209.65.80.66 (talk) 16:37, 4 February 2011 (UTC)

Additional stats

Should there be articles for additional browser statistics such as:

• Usage share of operating systems of web browsers
• Usage share of display resolution of web browsers
• Usage share of display resolution of web browsers
• Usage share of color depth of web browsers (Though virtually all are now 32bit)

Smallman12q (talk) 18:47, 4 February 2011 (UTC)

January wikimedia update

On Sun, Feb 6, 2011 at 7:48 AM, Erik Zachte <ezachte@....org> wrote:

On 2/6/2011 1:34 AM, Daniel Cardenas wrote:

Hi Eric,

Do you know if there are instructions somewhere, on how to create the

traffic analysis report? I'm interested in seeing the data for January.

http://stats.wikimedia.org/archive/squid_reports/2010-12/SquidReportClients.htm

Thanks,

Daniel

Hi Daniel,

I will update the report in coming week when I get home from US conference. Complete automation of the reports is on the to do list, but that may take a while.

Best, Erik
Daniel.Cardenas (talk) 18:02, 6 February 2011 (UTC)

old stat pages

copied from archive 2:

== World Wide Web Survey 1994-1998 ==

How come the other browsers in the GVU WWW user survey (January 1994 to October 1998) aren't listed? There are also other surveys at [4].Smallman12q (talk) 14:02, 5 June 2010 (UTC)

I found some more interesting historic links:

• CNN - Behind the numbers: Browser market share
• UIUC stats

mabdul 14:18, 14 February 2011 (UTC)

Add w3schools

I was thinking of adding w3schools with disclaimers. Yes, it is not representative of global usage share but neither is statowl and is a popular site in general and for browser stats.   http://w3schools.com/browsers/browsers_stats.asp   What do you think?   Thanks, Daniel.Cardenas (talk) 19:06, 29 January 2011 (UTC)

I think we've had this discussion before (check the archives). The reason for not adding w3schools boils down to: The stats are only for their site, unlike the other sources which use many diverse sites. This is further compounded by the specific nature of the site (web development) which is a very niche market. Now, if the other editors agree to adding w3schools, I might be willing to bend, but the stats really are an order of magnitude divergent from the median. Jdm64 (talk) —Preceding undated comment added 22:55, 31 January 2011 (UTC).

For the reasons Jdm64 listed above, and per previous discussions - I am against adding w3schools to this article. Wikiolap (talk) 03:37, 1 February 2011 (UTC)

The stats do not indicate an order of magnitude different. So you feel a little bit whacked like statowl is o.k. but more whacked is bad? Kind of playing favorites don't you think? Daniel.Cardenas (talk) 05:44, 1 February 2011 (UTC)

Well, what would be your definition of order of magnitude different? Because I'd say that the inverted stats between I.E. and Firefox is abnormally different, especially since the median difference between the two is about 15%. But if you don't like that reason, look at it given the fact that all the other sources gather data from multiple domains, except w3schools. Jdm64 (talk) 09:43, 1 February 2011 (UTC)

and except wikimedia. But I have to say that I'm also against adding it. The discussion was quite clear: web developers/technic-affines are choosing software that their needs serve; others often don't know that they have a choice. For that reason the EU created browserchoice.eu! mabdul 10:23, 1 February 2011 (UTC)

To answer Daniel.Cardenas question about playing favorites - the criteria I try to judge this is outlined above - if source aggregates wide range of sites - it is OK, if it reports only for its own site - then it is not OK. This seem like a clear and reasonable criteria, and it is accepted by most of the editors here. "degree of whackiness", on the other hand, is very subjective criteria :) Wikiolap (talk) 17:15, 1 February 2011 (UTC)

Google is a single site, Wikipedia is a single site and they are not an aggregate of wide range of site. --Sandro kensan (talk) 20:32, 1 February 2011 (UTC)

Wide is a subjective term. Is browser statistics based on 92% of one country wide?

http://en.wikipedia.org/wiki/Order_of_magnitude - "In its most common usage, the amount being scaled is 10 and the scale is the (base 10) exponent being applied to this amount ..."

15% ? You should remove statowl also then.

Daniel.Cardenas (talk) 21:02, 1 February 2011 (UTC)

A few points. (a) We have Wikimedia, not Wikipedia stats here. (b) It's not about how many sites, it's about how representative a sample the figures are of all actual web users. (c) Figures from one site that serves mainly people who want to learn to build websites, or from several sites that all serve mainly US American users, are not representative figures of the World Wide Web (d) Even if we did use just Wikipedia figures from every language (which we don't, see (a)), that would still be a very representative sample of worldwide usage. As we include all the other Wikimedia projects worldwide too, it's even bigger and better. I am against using small, narrow and non-representative samples (especially ones that provide mostly outliers in each stat, or other anomalous results), but I don't see Wikimedia's figures in that category at all. --Nigelj (talk) 21:36, 1 February 2011 (UTC)

How do you feel about statowl? Is that too small and narrow? Does w3schools compensate well? I think if we are going to allow lopsided stats than go ahead and allow more of them. Daniel.Cardenas (talk) 23:22, 19 February 2011 (UTC)

If we had thousands of skewed data sets, I'd agree with you - add them all - but we only have two or three such data sets. If we add so few, we just end up looking at noise instead of signal. Therefore, my view is that we should exclude any data that does not (as best as we can tell) represent a broad cross section of the actual web users worldwide. When I say exclude, I don't mean from the page or from discussion or consideration on the page, but from the tables of summary stats that are used to come up with our headline figures (the medians). I think we have a responsibility here as an important tertiary source - where else are people going to look to find an overview answer to this question? Wikipedia is highly rated in every way, and our overview/headline carries some weight. --Nigelj (talk) 10:28, 20 February 2011 (UTC)

My point is lets force the issue to get a better wikipedia. Lets add the skewed data, so people can't pick and choose their skewed data. Either its all in there or its not. Therefore we will get a better wikipedia. Do you think wikipedia is better having select skewed input? Daniel.Cardenas (talk) 18:12, 20 February 2011 (UTC)

Chrome in Albania

Chrome has already become the most popular browser in Albania with a marketshare of 50% and over. —Preceding unsigned comment added by 95.232.87.60 (talk) 15:14, 21 February 2011 (UTC)

You can't take a median of percentages!

It's mathematically incoherent to take a median of percentage values, as the chart at the top of the article does (or did -- I just removed that row). There is absolutely no reason that the medians of a set of percentage data would sum to 100%, because unlike the mean, the median does not have the distributive property. Consider:

Data points:

Foo   Bar   Baz | SUM
15%   25%   60% | 100%
40%   10%   50% | 100%
 5%    5%   90% | 100%

Median:

15%   10%   60% | 85%

Oops! We got values that summed up to 85%, that can't be right! So I've removed the median row from the chart at the top. --FOo (talk) 06:07, 22 November 2010 (UTC)

We aren't concerned if they add up to 100%. The numbers to start with don't add up to 100% anyways. Look at the numbers for w3counter for example.   Daniel.Cardenas (talk) 06:45, 22 November 2010 (UTC)

I agree that the median row in the table is valid. The thing that wouldn't be valid would be to display the median row as a pie chart, as, e.g., 49% could well look like much more (or less than) than half. The way to display the set of medians would be as a bar graph as we did for a while, but we don't do that at the moment anyway. --Nigelj (talk) 09:56, 22 November 2010 (UTC)

I don't strongly agree with the pie chart but agree anyways. Next month I'll make it a bar, if someone doesn't do it sooner. By the way the current median adds up to 100.6% , so 49% should look like 49%. :)   Daniel.Cardenas (talk) 15:30, 22 November 2010 (UTC)

Can I add back the original char? And what median do you want it based on? Both tables are within less than 0.36% difference. Jdm64 (talk) 22:03, 22 November 2010 (UTC)

Please use the world-wide table. Thanks, Daniel.Cardenas (talk) 01:06, 23 November 2010 (UTC)

Yes - please bring back the original chart. I beleive using median from summary table is better - please use it. Wikiolap (talk) 02:06, 23 November 2010 (UTC)

Ok, original chart back, using the world-median. But, like I said before the difference between the two tables is less than 0.36%. So, I think that shows that there's really no need for the duplicated world table. Jdm64 (talk) 02:40, 23 November 2010 (UTC)

It's still the case that a median of percentages is of extremely limited significance, and is likely to present a distorted view of the data.

The reason the rows do not sum to 100% is that some data have been excluded in this particular summary presentation (namely, the "other" browsers); obviously, the actual percentages across the full set of data would sum to 100%. (If not, they aren't percentages at all.) That's not a distortion of the data; it's just not presenting the least interesting data.

However, taking the median of percentages from different surveys does distort the data, as in the example I gave above. There is no reason to expect it not to do so. A median is typically used to express the central value of a statistical population (e.g. median age; median income), not to summarize across multiple summary values.

I'll go and seek additional mathematical advice on this subject; watch this space. :) --FOo (talk) 14:58, 5 December 2010 (UTC) It is a mean not a median. 91.109.159.10 (talk) 19:34, 22 February 2011 (UTC)

to do: expand wikimedia mobile browsers

Mobile Safari at 3.73% + desktop at 5.53% = 9.26%. That's an interesting statistic. :) I was thinking of having sub headings. Safari with 3 subheadings of desktop, mobile, and total. Don't know if we should do something similar for mobile other. Such as subheadings for android, ...
Disappointing that statcounter stats don't come close.   Daniel.Cardenas (talk) 17:17, 15 February 2011 (UTC)

Usage share data from Wikimedia visitor log analysis report

Period	Internet Explorer	Firefox	Chrome	Safari			Opera	Mobile other
				Desktop	Mobile	Total
January 2011	41.56%	28.71%	11.75%	5.53%	3.73%	9.26 %	3.55%	3.17%

Probably need to have total mobile and non safari mobile. Kind of messy.  :(   I think I'm not going to bother with the whole thing, unless someone has a better idea.  :)   Daniel.Cardenas (talk) 06:38, 16 February 2011 (UTC)

what about opera and opera mobile, opera mini, opera wii and opera ds? mabdul 10:46, 16 February 2011 (UTC)

Usage share data from Wikimedia visitor log analysis report

Period	Internet Explorer	Firefox	Chrome	Safari			Opera			Android	Mobile
				Desktop	Mobile	Total	Desktop	Mobile	Total		Other	Total
January 2011	41.56%	28.71%	11.75%	5.53%	3.73%	9.26 %	3.55%	0.70%	4.25%	0.90%	1.57%	6.90%

Getting unwieldy don't you think? Daniel.Cardenas (talk) 17:40, 16 February 2011 (UTC)

I have a widescreen monitor... For me it doesn't matter... The question is should we open the box of Pandora? (what about mobile ie, chrome will also go mobile. FF is already mobile) mabdul 18:02, 16 February 2011 (UTC)

Another suggestion is to have two tables - for desktop browsers and for mobile browsers. It probably will make difference more visible, because market shares in mobile are very different from desktop...Wikiolap (talk) 05:50, 17 February 2011 (UTC)

I do also agree that this would be the better solution. mabdul 12:39, 17 February 2011 (UTC)

What do you think of 5.53+3.73=9.26? I think most people want to see that. How would you show it? Daniel.Cardenas (talk) 19:47, 17 February 2011 (UTC)

I know that the Firefox stats are wrong! (defacto) Shouldn't we add the Iceweasel stats to Firefox since they are IDENTICAL BROWSERS? IE mobile and FF mobile should get also part of this table then... As Wikiolap said: make this really sense to compare mobile with desktop versions? separate (additional) table with desktop against mobile browsing would be good like this: mabdul 20:11, 17 February 2011 (UTC)

Usage share data from Wikimedia visitor log analysis report

Period	desktop	Mobile
January 2011	93.1%	6.90%

Don't understand. Are we going to display the 9.6% for Safari?? Daniel.Cardenas (talk) 04:52, 18 February 2011 (UTC)

sry for the delay. I meant that we could create an extra table for displaying the percentages between the mobile and the desktop stats. this is or maybe this will become important to see the changing of the browsing behaviour. I won't update the charts with safari, but let us discuss this in the section below. mabdul 10:38, 27 February 2011 (UTC)

Unique users

What NetApplication means with unique users? From [5] there are 2 billion unique users in Internet, NetApplication count 160 million unique visitors (8% of the Net):

'Netapplications bases its usage share mainly on the usage statistics of 40,000 websites having around 160 million unique visitors.'

Is possible monitored 8% of the word wide web with only 40 thousand website?

The answer is NO. --Sandro kensan (talk) 20:49, 28 February 2011 (UTC)

Precision in summary table

Do we really need two decimal places for the majority of the figures in the summary table? Looking at the spread, we can see that the last figure at least is all but meaningless. The fact that some figures are to only one decimal place makes the table harder to scan. It would be easy to reduce all the figures to one decimal place (rounding up or down appropriately), but I don't want to do it if it will only upset people. Everything else is looking very good at the moment IMHO. --Nigelj (talk) 22:10, 7 March 2011 (UTC)

Agree. Thanks, Daniel.Cardenas (talk) 22:19, 8 March 2011 (UTC)

Statcounter mobile browser vs platform

Wikimedia lists safari having much higher usage than opera. Statcounter lists platform iphone and ipod touch.[6] Would be nice if one could assume this was safari mobile, but from the info available, we can't make this assumption? Adding those two together gives a number that correlates better with wikimedia usage. Daniel.Cardenas (talk) 06:36, 9 March 2011 (UTC)

Historical Usage Share chart oddity

I like the line charts in the "Historical usage share" section (2 for Stat Counter, 1 Stat Counter Europe, 1 Net Applications). However taking a close look I notice the y-axis value starts below 0, with 0 being slightly above the x-axis. This gives the appearance of little relative growth, for quickly growing lines near the bottom. For example a browser that grew from 0% to 3% market share between 2010 and 2011 has a line along 0% from 2008 to 2009, and with the 0% significantly above the x-axis, a quick glance without reading the y-axis labels gives the appearance of this hypothetical browser having little relative growth 2008 to 2011, and with market share predating 2007. I'm certainly no expert at making charts, but I wonder if the y-axis can get shifted down to put 0% at the start? Or if this muddies the x-axis line too much, can another line be added to mark the 0% baseline? It's strange to have a blank area representing negative market share. Jeffhoy (talk) 15:27, 15 March 2011 (UTC)

I'll try to remember to look what R feature will fix that, when I recreate the graphs. Daniel.Cardenas (talk)

Note about median value not overly mathematically sound

Reference this previous discussion: http://en.wikipedia.org/wiki/Talk:Usage_share_of_web_browsers#You_can.27t_take_a_median_of_percentages.21

And this edit: http://en.wikipedia.org/w/index.php?title=Usage_share_of_web_browsers&action=historysubmit&diff=421550194&oldid=421537583

I'm thinking it would be good to add a note that says something like the median value of a set of percentages is a bit whacked, but we do it anyways. I've seen references that discuss this also. What do you think? Thanks, Daniel.Cardenas (talk) 01:23, 31 March 2011 (UTC)

Other Graphs

I found this website, which says IE has 54.27% of market share. http://marketshare.hitslink.com/browser-market-share.aspx?spider=1&qprid=0 Some unusual browsers though. — Preceding unsigned comment added by P3771 (talk • contribs) 18:51, 4 June 2011 (UTC)

That is Net Applications, they are referred to extensively. Thanks Daniel.Cardenas (talk) 21:30, 5 June 2011 (UTC)

I'm guessing that the Playstation browser is not included in the table due to its not in other tables?P3771 (talk) 14:13, 6 June 2011 (UTC)

No, it isn't inlcuded since it has only such a low market share. mabdul 14:23, 6 June 2011 (UTC)

Problem: thumbnails not getting updated

I updated many of the graphs but the updates aren't showing up due to system issue. If you click on a chart all the way to the end you will see the update. Daniel.Cardenas (talk) 13:30, 17 July 2011 (UTC)

Known problem. Techs are working on it. (at least for a week or so). mabdul 17:21, 17 July 2011 (UTC)

I updated them manually. Daniel.Cardenas (talk) 22:04, 17 July 2011 (UTC)

World Map

Do we have any references for the world map? While it's a neat idea, I question its accuracy - it lists North Korea as predominantly using Internet Explorer, despite the fact that the country has virtually no web browsing statistics to mine, and considering the Red Star Operating System is UNIX-based, even if there were (or even if we were counting Kwangmyong statistics), I'd be inclined to think Firefox would predominate (and even if it didn't, the lack of a functioning IE for UNIX would make me think that at the very least, it would not be IE). 75.154.124.62 (talk) 19:05, 11 May 2011 (UTC)

Please use proper citations for the browser world map. It was originally created by browserrank (at www.browserrank.com) in January 2011 and has been updated regularly since. At the time I proposed including that cite here. Now I see months later it is attributed to someone else. — Preceding unsigned comment added by 99.142.79.171 (talk) 16:10, 9 August 2011 (UTC)

how the stats are collected? and discrepancy

50 to 30% is a HUGE difference. There is nothing in the article about the sites' sampling, how there could be a large discrepancy, etc. Are we supposed to accept the stats, hook, line, and sinker, or can someone write a little something about how the numbers can differ, what the sample sizes are, etc. Is the sample size near the size of the internet, or is it smaller? Some of the article addresses that, for example the Wikimedia, but not all of them have accountability listed with them. Please fix it.

--Hitsuji Kinno (talk)

Added a section: "Differences in measuring".   Feel free to improve.   Daniel.Cardenas (talk) 20:56, 8 August 2011 (UTC)

I removed the statement that weighted statistics cause areas that have a higher number of Internet users to be weighted more in the stats. They don't. They compensate for the fact that the stats collected are biased towards certain geographical areas. For example, NetApplications does not get collect much data from China. They compensate for this fact by weighted the data they have collected from China higher, correcting the problem with a smaller amount of data collected from that region. This makes the normalized statistics less biased, not more biased as the article seemed to imply. -- Schapel (talk) 22:32, 8 August 2011 (UTC)

Discussion on Narrow Stats

Statowl represents the U.S. mostly and w3schools represents web technology enthusiasts. Should we have these narrow stats in the summary table? Previous discussion suggests that Statowl is not too narrow, while w3schools is too narrow. Please share your thoughts below. Perhaps the most relevant wikipedia policy is Neutral point of view. I suggest either both or neither be allowed and not pick and choose your favorite statistic. Daniel.Cardenas (talk) 20:50, 26 February 2011 (UTC)

statowl represents stats from many different sites while w3schools only shows stats for its own site - this is the reason to have statowl stats and not w3schools. This is also the consensus that was previously reached on this page. Until different consensus is reached, please respect the previous decision and don't make new changes to the article in this area. Thanks ! Wikiolap (talk) 05:40, 27 February 2011 (UTC)

Instead of defining it multi sites versus single site, how about global reach versus primarily regional stats? Daniel.Cardenas (talk) 18:44, 28 February 2011 (UTC)

Remove both.--Sandro kensan (talk) 20:41, 28 February 2011 (UTC)

The question about statowl has been discussed before - see the discussion and votes here: Talk:Usage_share_of_web_browsers/Archive_3#Remove_Statowl_from_summary. Likewise, w3schools has been also discussed before - Talk:Usage_share_of_web_browsers#Add_w3schools. Both discussions represent current consensus. Consensus can be changed, of course, you can request new vote (and keep rerequesting it every month) - there is nothing in the code of Wikipedia which would prevent that. Even though there don't seem to be new arguments, the corpus of editors change, their opinions change, and the result of vote may change too. (I personally find the jihad for removal of statowl from summary table a bit misplaced, and tactics used being questionable). Wikiolap (talk) 06:44, 1 March 2011 (UTC)

Yes please cease your jihad and your questionable tactics. Claiming consensus to keep statowl when the consensus was to remove it. Please don't add it back in, until the consensus changes. Daniel.Cardenas (talk) 17:15, 1 March 2011 (UTC)

Daniel - I apologize for poor choice of words - it was unprofessional. I try not to trick the system, and I should assume that other editors act in good faith as well.Wikiolap (talk) 15:16, 2 March 2011 (UTC)

I appreciate your contributions. Thank you! Daniel.Cardenas (talk) 19:14, 2 March 2011 (UTC)

I don't understand the problem right now: what changed in the last few months? mabdul 18:37, 1 March 2011 (UTC)

There are/were 6 remove votes for statowl and 3 keeps. Seems the keep camp incorrectly believed they had a consensus to keep. Daniel.Cardenas (talk) 21:30, 1 March 2011 (UTC)

And I didn't even voted... (oh and I would vote against!) mabdul 22:31, 1 March 2011 (UTC)

In light of the previous consensus linked above, as well as my own opinion on the matter, neither statowl nor w3schools should appear on the list. If either of them are currently there, they should be removed. TechnoSymbiosis (talk) 23:16, 1 March 2011 (UTC)

Now after the removal of statowl from the summary table, the summary table in the middle of the article and worldwide view table at the beginning became exactly the same. Is there a reason to keep both of them now ? I propose that one of them is removed (I don't have an opinion about which one). Wikiolap (talk) 05:55, 2 March 2011 (UTC)

I'm not a contributor to Wikipedia, i was just browsing to find web browser statistics. I was horrified to see w3schools statistics and the link to w3schools site in this article. I was so horrified, that i had to register and complain :) Many people think of w3schools as a legitimate resource for web development, but in fact it is not. W3Schools is a commercial site with a very aggressive SEO marketing, and I'm afraid that linking them in wikipedia articles might be part of their marketing strategy. Learn more about w3schools here: http://w3fools.com/ I really don't see the reason why their totally non-representative data is included, when there are multiple, more reliable sources. I strongly argue for the removal of the w3schools data, and any links to their site. Babett (talk) 19:20, 9 April 2011 (UTC)

I agree. Many people refer to the W3Schools stats as if they're representative of overall web usage, but it's one of the most biased samples I know of. There are plenty of other more representative samples available. I think we have plenty of samples already and can simply remove W3Schools stats. -- Schapel (talk) 21:31, 9 April 2011 (UTC)

I Don't Count. Since I am the only developer working on statowl.com - I am biased - however, I don't think painting our data as "narrow" is entirely accurate. As I pointed out here Talk:Usage_share_of_web_browsers/Archive_3#Remove_Statowl_from_summary, we are not just US data. We also source data from hundreds of different sites which do not target users with specific technological interests (i.e. stats gathered from a blog about fedora linux topics would bring highly skewed data so we would not track that site). We are also continually expanding our data sources outside the US, but as things stand now, around 80% of our data is from US traffic. Any assistance in expanding our non-US data sources is greatly appreciated. We will be tracking a lot more Spanish content later this year and I hope this expands our footprint in countries where Spanish is the predominate language. Mikemc100 (talk) 05:39, 16 May 2011 (UTC)

Perhaps in a year or two if the samples on statowl.com spread more it would be welcomed but as of now it is quite clearly not fit for statistical use on this page. — Preceding unsigned comment added by 72.241.135.189 (talk) 16:43, 31 August 2011 (UTC)

It looks like there was agreement that StatOwl should be in the summary table, but one editor removed it. I'm not sure why. I suppose the argument is that because most of StatOwl's data is from the U.S., it's biased. But all of the sources are biased in one way or another. The bulk of NetApplication's data is from large companies, causing it to report a large figure for IE usage. W3Counter counts only a certain number of page views from each site, which skews its data towards browsers used on small sites, and we can see that it reports a small figure for IE usage. Wikimedia attracts most page hits from young technology enthusiasts, so again its data shows a low figure for IE. I think we should include StatOwl and Clicky in the summary table, so all the recent sources are summarized in the table. In any case, the most biased numbers will not affect the median much, because the median represents what's left after eliminating the most biased figures. -- Schapel (talk) 19:23, 31 August 2011 (UTC)

What makes you say: "It looks like there was agreement that StatOwl should be in the summary table..." ? Daniel.Cardenas (talk) 19:38, 31 August 2011 (UTC)

The comment on this edit. -- Schapel (talk) 21:59, 1 September 2011 (UTC)

See this comment above:

There are/were 6 remove votes for statowl and 3 keeps. Seems the keep camp incorrectly believed they had a consensus to keep. Daniel.Cardenas (talk) 21:30, 1 March 2011 (UTC)

Daniel.Cardenas (talk) 03:55, 2 September 2011 (UTC)

Browser version (IE6, IE7, IE8, IE9)

Especially for Internet Explorer, it is useful to know the market share of specific versions. Web developers often need to decide if it is worth to support IE6 or not. I think this article should either provide those version-specific stats directly, or link to a place where you can find them. -- 78.48.230.55 (talk) 15:25, 22 June 2011 (UTC)

This article contains many links to find various stats for different versions of browsers. However, a web developer would be foolhardy to rely on those stats to decide whether to support a particular version of a browser on a particular site. Web developers should rely on the stats for each individual site, because they can vary wildly from site to site. -- Schapel (talk) 02:23, 23 June 2011 (UTC)

Does not mean such stats are not useful for a web developer. Individual site stats are not always an option, especially on smaller sites, and pre-launch. Anyway, if there are some links that is sufficient. Unfortunately, the text is so long that it's hard to look for such links specifically. E.g., I tried to find-in-page for "version", and did not find what I was looking for. I remember this article does have such links, I am just not able to find them. So what about a section with links to version-specific stats, that is easy to find. (sorry for not doing it myself, but I am soo not into this article, it would take me hours to even get started) --78.48.37.118 (talk) 19:43, 25 July 2011 (UTC)

This article stinks of bias towards stats companies - it doesn't describe progressions or factors behind usage share but merely uses stats tables from commercial organisations. Really needs some 'Quality Control'! — Preceding unsigned comment added by 124.169.29.58 (talk) 07:55, 5 August 2011 (UTC)

In the past when author(s) tried to add progressions or factors other author(s) said that text belongs in browser wars article. Daniel.Cardenas (talk) 19:28, 13 August 2011 (UTC)

I agree that version information is the thing that would be really useful. I don't care at all if IE is used more than FF. Is anyone still using IE5.0 or FF3.6? — Preceding unsigned comment added by 124.149.99.98 (talk) 03:21, 13 August 2011 (UTC)

That information available through the following links: Firefox, Internet Explorer. You can find even more detailed information by following the links to the sources in this article. That's the beauty of a well-linked, well-sourced wiki -- you can follow the links to get other information. -- Schapel (talk) 15:10, 13 August 2011 (UTC)

Adding mobile browser to the main browser for safari and opera

For statcounter, was thinking of adding mobile browser to the main browser for safari and opera. This has been done for wikimedia and with opera for netapp. A problem that I see with this is that for statcounter top mobile browsers it doesn't say safari, instead it says iphone and ipod touch. http://gs.statcounter.com/#mobile_browser-ww-monthly-201108-201108-bar . Would it be wp:synth to interpret these as safari browser?   Thanks, Daniel.Cardenas (talk) 19:31, 20 August 2011 (UTC)

I'm fine with adding up the iPad and iPhone users to the total Safari users. We should keep Opera and Opera Mini separate, because those are separate products. -- Schapel (talk) 17:41, 26 August 2011 (UTC)

I was suggesting adding safari and safari mobile. Those are separate but similar products. Are you o.k. with that? Daniel.Cardenas (talk) 17:53, 26 August 2011 (UTC)

Others may disagree, but I would say that Safari and Safari mobile are different versions of the same product, Opera and Opera mobile and different versions of the same product, and Firefox and Fennec are different versions of the same product. Opera and Opera Mini, on the other hand, are completely different beasts. Opera Mini is a Java program that requests web pages through Opera's servers. It's not really even technically a web browser, as it's incapable of browsing the web on its own. -- Schapel (talk) 23:23, 26 August 2011 (UTC)

Added up Iphone and Ipod numbers and created a new table. Looks ugly. Let us know if you have suggestions.   Daniel.Cardenas (talk) 04:12, 3 September 2011 (UTC)

Perhaps having a single column of Safari and having a note that it includes iphone and ipod would be better. Daniel.Cardenas (talk) 04:20, 3 September 2011 (UTC)

Opera and Opera Mini

The Opera Mini article says that it is derived from Opera (web browser). So they are related, even though distinct products. Daniel.Cardenas (talk) 23:07, 3 September 2011 (UTC)