Stuck With Old Browsers for at Least 3 Years

by Jakob Nielsen on April 18, 1999

Continued conservatism of Web users means that:

  • sites will have to support users with version 3 browsers until early 2001
  • version 4 will have to be supported until late 2003
  • Netscape 5 will never get more users than Netscape 4
  • not until the year 2003 will advanced browsers be sufficiently widely used to allow sites to go beyond the basics

In March 1998, I wrote an Alertbox commenting on the reluctance of Web users to embrace new technology. I noted that the speed with which users upgraded from Netscape 3 to Netscape 4 was about half of that seen for earlier browser releases.

Current data on the use of different Netscape versions from StatMarket shows a remarkable continuation of the assessment I made more than a year ago. Plotting the new data on a chart with the old trend curves (see below) results in utterly straight lines. I don't mind being proven right, but it's uncanny to be this right.

Chart showing how many users have different versions of Netscape
Use of various versions of Netscape from late 1995 to January 2002.
Heavy lines show empirical data; thin lines predicted data.
Sources: Interse (1995-1997), AdKnowledge (1998), StatMarket (1999)


The chart also shows predicted upgrade curves for the next few years, including a hypothetical Netscape 6.0 release at the end of the Year 2000. I have assumed that future upgrade curves follow the trends we have observed recently, though it is also possible that they will slow down even further. After all, the more mainstream users come online, the less people will obsess about having the absolutely latest software installed.

A particularly interesting aspect of the predictions is that Netscape 5 is a "lost generation" that never rises above Version 4 on the chart. In fact, it may never again be possible for a single browser version to dominate the Web the way Mosaic and Netscape versions 1-3 did. Even Netscape 4 didn't truly dominate the way the chart shows because it had to share the limelight with IE 4.

There is less data available about the upgrade curves for Internet Explorer, but the evidence seems to suggest that Microsoft has slightly more success in getting new versions used; maybe because they ship them with the operating system. Even if IE has a slightly faster upgrade speed than Netscape, the conclusion remains the same: IE 5 will also not dominate the Web. Instead, 1999 and 2000 will see a proliferation of new devices used to access the Web, from WebTV over Internet mobile phones to new types of information appliances.

Back-to-Basics Web Design

Microsoft has been justly criticized for several standards deviations in Internet Explorer 5.0. It is indeed amazing that Microsoft can't implement two-years-old specifications correctly, but in fairness it must be said that IE 5 is the best browser yet and that it does support almost everything in the basic Web standards (HTML 4 and CSS 1).

Ultimately, the weaknesses in IE 5 don't matter much since websites are not going to be able to implement any advanced features for several years to come. Once versions 6 and beyond take over in the Year 2003, new designs will become possible, but for the next four years it will be better to focus on improving content, information architecture, navigation, and search; all of which are independent of the browser.

The slow uptake speeds and the bugs and inconsistencies in advanced browser features constitute a cloud with a distinct silver lining: Recognizing that we are stuck with old technology for some time frees sites from being consumed by technology considerations and focuses them on content, customer service, and usability. Back to basics indeed: that's what sells since that's what users want.

Update 2003

As of February 2003, the marketshare for Netscape 4 was 1.1% and IE 4.0 was at 0.9%. In total, 2% of Web users were still on v.4 browsers .

It's hard to say whether it is acceptable to ignore 2% of the customers, but mainly I would say that my prediction from 1999 turned out to be pretty accurate, though not 100% correct:

  • until now, it has continued to be necessary to support "old browsers" (by which I meant v.4 or older)
  • as of early 2003, most sites can relax this restriction (instead of waiting until late 2003, as originally predicted)

Of course, it's still necessary to support v.5 browsers.

One point where this article was simultaneously proven right and wrong was in discussing the fate of a hypothetical version 5 of Netscape. I predicted that this version would never amount to anything great -- instead what happened was that Netscape cancelled this release altogether. Hard to blame them given the gloomy prospects I had predicted.

Internet Explorer Upgrade Speed

IE v.6 was released in August 2001 and by February 2003 enjoyed a market share of 58%. In other words, the upgrade speed was 0.8% per week.

This is slightly less than the 1% per week upgrade speed for Netscape in 1998.

Update 2008: Uptake Speed Cut in Half released statistics for the browsers used to access their service in June 2008. This is much more useful than the usual browser stats, because users represent enterprise customers and thus the type of users likely to patronize B2B sites, enterprise services, and other high-value uses of the Web.

83% Internet Explorer
14% Firefox 0% Safari 3% Other

(Safari was actually at 0.4%, which rounds to 0%.)

Of IE users, 38% use IE7 and 62% are still using IE6.

Remember that IE7 was released in October 2006: 20 months before this data was collected. Thus, the uptake speed was slightly less than 2% per month (in terms of IE users upgrading from the old version to the new one).

Ten years ago, in 1998, I noted that the uptake speed for new browser versions was 1% per week, which equates to about 4%/month. Thus, users are now twice as conservative as they used to be (upgrading at half the previous pace).

The guideline to stay at least one version behind on employing any new browser or plug-in technology is even more relevant than it ever was. If we take 2%/month as a rough rule of thumb, it takes slightly more than 3 years for any new release to reach 3/4 penetration, which is the absolute minimum for you to even consider using it.

Share this article: Twitter | LinkedIn | Google+ | Email