Jakob Nielsen's Alertbox for October 18, 1998:

Failure of Corporate Websites

On the average, the Web doesn't work: when you think of something to do on the Web, the expected outcome is that you will fail. Some recent data to support this claim:

In Jared Spool's study of 15 large commercial sites users could only find information 42% of the time even though they were taken to the correct home page before they were given the test tasks
A study from Zona Research found that 62% of Web shoppers have given up looking for the item they wanted to buy online (and 20% had given up more than three times during a two-month period)
Forrester Research audited 20 major sites, finding 51% compliance with simple Web usability principles such as "is the site organized by user goals?" and "does a search list retrievals in order of relevance?" (in other words, the average site violated half of these simple design principles)

Despite these miserable statistics, users do benefit from the Web since they spend most of their time on the good sites. But the odds are against them when they want to try something new. And the odds are against any company that wants to put up a website: in my estimate 90% of commercial websites have poor usability.

The recent Forrester report is particularly interesting because it tries to identify the reasons for the many bad corporate websites as well as the impact on a company from having a bad site. Many of Forrester's conclusions are similar to my writings in the Alertbox since 1995 and the report provides additional supportive data from large corporate Web projects.

For the report, Forrester interviewed 25 executives in charge of various companies' Internet efforts. Most had very few design goals for the site, though 56% did mention "fast performance" as a goal (bravo!). 24% of sites conducted usability testing: this is more than I would have expected and probably reflects Forrester's bias toward bigger and more well-funded projects. (Of course, this data implies that 3/4 of large sites are managed without any usability data: essentially poking blindly into the design space.)

Impact on the Company

Forrester estimates several costs of bad Web design. The two most striking are:

loss of approximately 50% of the potential sales from the site as people can't find stuff
losing repeat visits from 40% of the users who do not return to a site when their first visit resulted in a negative experience

The report provides considerable detail on the economic impact of these and other problems. In summary: many millions of dollars lost for a large website.

Fixing the Problem

The most dramatic fix suggested in the Forrester report is to close down sites that are so bad that they damage the reputation of the company. In most cases, various improvements are possible, at costs that are estimated from a few thousands dollars for removing linkrot to half a million for advanced solutions. Even the high-end solutions are cheap compared to the estimated cost of having a bad site.

The report predicts that Web solutions agencies will soon include usability engineering as one of their core offerings. I definitely agree, though I also think that the client needs to take an active role in auditing the usability of the deliverables.

Clients should insist on usability testing of all Web designs: The quality of the proposed usability process is one of the few ways a client can judge the quality of the end result while still in the proposal stage. A proposal without usability engineering milestones (or with poorly defined or misguided methodology) will result in a poor site most of the time (unless you are in the lucky 10%).

Reference

Harley Manning, John C. McCarthy, and Randy K. Souza: Why Most Web Sites Fail, Interactive Technology Series, Volume 3, Number 7, Forrester Research: September 1998.

If you work at a big company that subscribes to Forrester's service, I strongly recommend that you get a copy of this report. It will be very useful in avoiding disasters in future Web projects.

Previous: October 4, 1998: Personalization
Next: November 1, 1998: Why Yahoo is good (but may get worse)

See Also: List of other Alertbox columns