From CSS Discuss
Shares in the browser market
Only you can decide (for yourself or on behalf of your client) what percentage of users you are prepared to offer a lesser experience (which may mean graceful degradation or - ultimately - a site which is impossible to use). If a site is perceptibly deficient in some way, it's very likely that the user will blame the site and not his or her browser (unless they realise and accept that their browser is elderly). So any perceived deficiency may reflect adversely on the organisation (or individual) whose site it is.
Perhaps the question to ask is not "am I happy to cater for 95% of visitors?" but "am I prepared to risk antagonising or even losing 5% of visitors (potential customers)?". If the cost of catering for a browser with, say, 2% market share is less than the likely value of the business (or goodwill or support) its users will bring in, then your strategy is obvious.
The best source of browser usage is your own site's log files, provided you have access to them and can analyse the browser strings accurately. Accurate analysis isn't easy since some browsers pretend to be others in an attempt to overcome defective "browser sniffing" by Web sites. (In fact, even IE6 pretends to be the non-existent Mozilla 5.) Also, an analysis of browsers should exclude other user agents such as spiders - and they are certainly not always easy to identify.
The free and widely used log analysis tool Analog gives a correct analysis of Opera usage. Version 6.0 onwards distinguishes Firefox from other Gecko-based browsers (such as Netscape and Mozilla).
AWStats is a decent free perl-based log analyzer with a lot better browser breakdown than most. It distinguishes properly between versions of IE and Netscape, and also can detect Safari, Opera, Mozilla, Firefox, Galeon and Konqueror as well as Phone/PDA browsers and many others. It also does a good job of separating out and analyzing robot/spider traffic, and gives some other interesting metrics that are not always found in free packages.
Publicly available stats
For sources on the Web, there is the main Stats page on Charles Upsdell's Browser News.
There is also the Global Statistics pages of The Counter (unfortunately, these stats weren't updated between May 2003 and January 2004). We don't know how accurate The Counter's browser identification is. For example, Opera's share reportedly dropped from January 2004 onwards from around 2% to 0.5%. So it looks as though it now includes only accesses where Opera claims to be Opera. Note that the installed default setting up to Opera 8 was to identify itself as IE6. Since January 2004 month-on-month trends for all browsers have appeared to be unnaturally erratic. The Counter's sample may not be representative of your country (or of your particular site). If you can't analyze your own site logs, The Counter's figures will give a very rough idea of global browser usage and trends - don't treat them as definitive.
[The Counter can only gain statistics from pages which use its counter code. Not everyone uses these counters. My site, and all the ones I've built in the last 5 years or so, don't use public counters fullstop: so all my visitors don't count towards The Counter stats. So, if you make sites for the same sorts of people I make sites for, don't expect the counter to be representative. - DougBTX]
If you're interested how browsers stats look in Poland, take a look at Ranking.pl, which monitors 7 million users [unknown contributor].
Finally, the stats on the W3 Schools site are sometimes mentioned but they refer to the W3 Schools site itself, whose audience will be very untypical of Web users generally and which is more likely to use alternative browsers to the mass-market IE.
To conclude, be wary of any stats which don't state their sample size and period and which don't provide information on how they identify browsers accurately. If that sounds like bad news, I'd suggest that you probably don't need to know a browser's share to the nearest 0.1% to decide whether and to what extent you will support it.
Contributed by Jim Wilkinson
An alternative answer
Hey there. Let me guess why you've come - you've come because you, as a responsible web designer, care about how your websites will appear in all the browsers out there. Today's web technologies - HTML and CSS - make possible some powerful and amazing designs, but let's face it, browser support is not all it could be. Netscape 4 couldn't render its way out of a box - let alone an inline. And IE/win couldn't render its way out of a box model. Opera? Page flow? Let's not go there, girlfriend. Even Gecko has its quirks - and with the AOL Gecko Integration That Is Foretold, you know that's a problem.
So what's a designer to do? Obviously, consult the browser stats! Oh, sites will claim to be the most accurate, and that guy on your mailing list will swear blind that he knows the most reliable source for browser stats, and the woman across the hall says at her alma mater they're still using Navigator gold 3.04. But they are all wrong - dead wrong.
Some of the best minds in web design have bent their mighty will toward the task and, time and time again, reached the same clearly true conclusions about the true nature of browser share. Here it is, folks - the only browser share statistics page you'll ever need.
- No, This One's The Best
- Every "browser share statistics" report is different
- But mine's better.
- NS4 is declining!
- NS4 is inclining!
- IE dominates
- NS4 is still a presence
- When AOL releases its Gecko-integrated client, the Earth's orbit will reverse
- But they won't let you turn off pop-ups
- What about Opera? Opera can pretend to be something else. What about Opera? Eh? Eh?!
- My browser share statistics can beat up your browser share statistics, nyah nyah nyah.
- 'Truly' accurate stats are actually impossible. Unless one goes to every 'net user in the world and checks his or her computer. You volunteering?
- It ultimately doesn't really matter
- You should really think about the browser shares of your intended audience (or existing audience) when determining how best to alter your standards-compliant code to work with various browser support
There you have it, my friends! Up to the minute statistics! Guaranteed true! Now, go forth and code!