Essays in this series about understanding your audience:
- Getting to Know Your Audience
- Categorizing Your Audience
- Identifying Important Operating Systems
- Identifying Important Browsers
Your customers’ experience of your site will be accomplished with, and mediated by, a browser application. Browsers parse and render HTML according to their own logic and rules, so not every browser will provide the same viewing experience of your site. In addition, there are no guarantees that all browsers will support or have enabled all of the functionality required by your site; for example, if your site requires that cookies be enabled, any browser that can’t handle cookies — whether through design or user choice — will not correctly handle the functionality. And furthermore, the quality of your code and its adherence to a correct DTD will also be reflected by a browser’s ability to render the code.
Just as it’s important to understand which operating systems your audience is using, so should you also understand the browsers and browsing applications
1. Find out what’s commonly available
The first place to look for browser information is at a listing of the myriad user-agents that have been released to the web. A user-agent is a browsing application — think a “software program” — that connects to a web server and accesses pages under the command of a user; user-agents include such applications as browsers, spiders, bots, proxies, special-needs browsers (like Braille readers and screen readers). I’d recommend exploring the issue of the kinds of agents available before digging into statistics on popularity. I’m amazed both at how many different agents are roaming the web, and at how varied are the agents used to access even a specialized site such as philosophe.com; I’ve posted a sample user-agent log from philosophe.com. A good place to find out about agents is [check link ] BrowserWatch’s Browser Boulevard.
The difficulty with user-agent logs is that the value assigned to an agent consists of whatever the programmer thought would be useful or appropriate: there are no standards for what information is included. Some of the data on agents may simply not be useful.
When you’re ready to look at browser statistics, check out one of the sites that publish statistics on browser usage. These statistics are generated from the sites’ server user-agent logs, and are useful because they provide a snapshot of a larger audience than any one web site’s logs are able to reflect. Generally, web sites that publish statistics for browser usage across a network of affiliated sites are the most useful, because the statistics won’t be as closely tied to the unique audience that one site might have; statistics from a network will provide a broader data sample. The following sites are examples of useful sources of statistics:
- [check link StatMarket.com]
- StatMarket has a wealth of useful information, and seems to be growing in member sites (that contribute statistics). I just wish their data were in tabular form so I could do my own number-crunching. See table 1 below for an example of the kinds of statistics available at StatMarket.com.
The following table shows StatMarket’s browser usage data for the last three month period (from reports generated on May 31, 1999), sorted in order of popularity. Only data for the ten most popular browsers is included.
TABLE 1: Browser Statistics from StatMarket.com, ranked by average popularity
Rank Browser Avg. Min. (Date) Max. (Date) 1. Microsoft Internet Explorer 4.* 52.78% 47.11% (5/19/99) 58.63% (2.21.99) 2. Netscape Navigator 4.* 25.62% 22.74% (5/31/99) 27.74% (3/17/99) 3. Microsoft Internet Explorer 5.* 6.99% 1.88% (3/18/99) 18.34% (5/31/99) 4. Microsoft Internet Explorer 3.* 6.38% 4.29% (5/31/99) 8.80% (1/13/99) 5. Netscape Navigator 3.* 5.39% 3.73% (5/31/99) 6.95% (1/12/99) 6. WebTV 1.80% 1.48% (3/17/99) 2.22% (1/13/99) 7. Microsoft Internet Explorer 0.* 0.80% 0.65% (4/22/99) 1.02% (2/21/99) 8. Netscape Navigator 5.* 0.09% 0.00% (3/2/99) 0.28% (5/16/99) 9. AT 3.* 0.03% 0.02% (5/31/99) 0.05% (1/21/99) 10. Netscape Navigator 2.* 0.01% 0.00% (5/31/99) 0.03% (2/16/99)
- [check link irt.org]
- I’m still trying to figure out who Irt.org is, but they do have an interesting chart of user-agents — interesting because some of these agents are new to me: what in the world is a “SpaceBison/0.01″?
- [check link Engineering Workstations WWW server]
- This is a specialized set of statistics from a university server at the University of Illinois at Urbana-Champaign.
- [check link University of Utah Health Sciences Center ]
- I stumbled across this site’s usage reports from a server log analysis tool run apparently as a chron job. I don’t necessarily recommend this data as being valuable, merely pointing out how this kind of information is available from a range of sources.
2. Factor in AOL
The statistics you’ll find in user-agent logs regarding AOL equipped visitors are misleading, because a.) AOL users connect to the web through a proxy server system that caches web pages according to a logic algorithm, and b.) AOL has a large family of browsers which are based on a number of different rendering engines. The result of these issues is that AOL statistics will be under-reported by your logs; AOL users may be viewing your pages without calling them from your servers, and any hits you do see may be recorded as hits by the user-agent their browser is built around.
AOL has a site designed for webmasters at [check link http://webmaster.info.aol.com/]; once there, click on the button labeled “Browser Detection” for more information on how to identify AOL browsers.
As far as using information about AOL browsers to decide which browsers are important to your audience, well, any engine used by a version of the AOL browser should be considered especially important. People who like AOL seem to really like AOL, and these folks may be likely to retain an older version of the browser. In addition, given AOL’s massive advertising campaigns and their free-disk mailings, many users new to the Internet are likely to try AOL’s service. If your audience is known to include web beginners, clearly AOL should rank high on your “must support these browsers” list.
3. Look at Your Site’s Logs
For most sites, the great majority of browser traffic is generated by a core set of browser models and versions: the browser market is pretty much owned by Microsoft and Netscape. It would be surprising if the bell-curve of browser stats was far off what you’d see on the sites mentioned above.
But your server user-agent logs can show you the boundary area unique to your site. You can find out which of the generally less popular browsers is favored by your audience. In fact, the popularity of some browsers might be very valuable data because it indicates an unrecognized mode for using your site. For instance, if you had a higher average for users of WebTV or browsers that run on palm-top computing platforms like Windows CE, then you’d realize that parts of your audience are not using the tradition computer platforms to shop your site; this translates to a new market. Likewise, data showing that you have concentrations of foreign language browsers indicates a potential to embrace an international market.
4. Analyze Your Statistics
Once you have several sources of browser statistics, you need to start making sense of the figures. You need to identify the trends of browser usage — especially as concerns your particular audience — and plug this new understanding into your design and test strategies. You should be able to provide answers to the following types of questions:
- What are the most commonly used browsers on the web?
- Which browsers will become more important as their popularity increases?
- Which browsers are favored by your audience?
- Are there sectors of your audience that aren’t being supported?
- What new browsers are looming on the horizon?
- What browsers are getting lots of attention from the media?
- Does your site require certain levels of security, for example SSL 3.0 or 128 bit encryption?
- Does these functional requirement mesh with the observed trends in browser usage?
A good place for browser statistics analysis is [check link http://members.home.net/hjhornbeck/trinkets/brs_info.htm]; H. J. Hornbeck has done the crunching for you, even if he can’t answer these questions for your site.
5. Monitor The Media
Mass media like newspapers, television, magazines, and yes, web sites, will give you a feel for which browsers are being showered with attention. For example, every time Microsoft or Netscape releases a new browser version, it seems like everywhere you turn you’ll find a review or a comparison shouting out the browser’s virtues or failings. Major positive coverage is an indication that a browser will increase in audience share; certainly, a new release by Microsoft or Netscape necessarily must be included in the set of browsers your site must accommodate.
With lesser-known browsers, media play is an indication that their audience share is likely to increase as people follow their curiosity. The point here is to not be taken by surprise. A strong quality assurance process should evaluate the site’s performance on every important browser, and sometimes it is potential that makes a browser important. Never be caught flat-footed by customer complaints about how the site doesn’t work on the brand X browser.
Some useful media sites on browser issues are:
- [check link InternetUser and CompuServe Present BrowserWars]
- [check link Cnet's topic center on browsers]
6. Make Some Judgments
Common design practice today is to create a site that’s highly visual and deterministic with regards to page format and layout. Simply put, this means that most sites violate their HTML DTD so severely that correct browser rendering of the page becomes a matter of shear chance with some of the lesser-known browsers. This failure is not the fault of these browsers, because they were designed to handle the correct DTD, but rather this the fault of designers who hack their HTML to work with the major browsers. This strategy discounts the concept that HTML should be similarly and appropriately renderable by any browser that supports the DTD.
In addition, don’t assume that a more recent version of a common browser will handle HTML exactly as its predecessors did. Bugs do happen, and in some cases the browser code is written with stricter adherence to the HTML specification, so bad but expected browser behavior may be removed.
So, putting aside my concerns that a site should be coded so as to support the widest possible range of browsers, my high-level recommendations for browsers every commerce site should work with is as follows:
- Netscape Navigator 2.*
- This browser is beginning to support tables, today’s mainstay for layout control.
- Netscape Navigator 3.*
- Navigator 3.04 set the standard for support of SSL 3.0. This is a great browser, and my favorite test browser.
- Netscape 4.*
- Too common to ignore. There are two series: 4.04 & 4.05 are popular, while 4.5* is the most popular of the Netscape browsers.
- Netscape 5.*
- Not very common, but still important.
- Internet Explorer 3.*
- Explorer 3.02 is the standard for this series for SSL 3.0 support. Still a common browser, especially on older corporate intranets. I hate this browser.
- Internet Explorer 4.*
- The most popular browser by most counts.
- Internet Explorer 5.*
- Newly released and growing steadily in popularity.
- AOL 3.*
- Extremely common. Most of the 3.* versions are built on the Internet Explorer rendering engine.
- AOL 4.*
- The most popular of the AOL browsers, most are built on the Internet Explorer 4.* engine.
- Still a small portion of the browsers in use, but growing in popularity.
7. Improve Your Quality Processes
As noted above, you should use your understanding of important browsers to make sure your site’s design and functionality accommodate your users. Your audience should essentially drive your design.
Include the most popular browsers in your testing, both for functionality and for testing the user experience. Combine your understanding of important operating systems and browsers to create a set of testing environments (what I call a test suite).