Jakob Nielsen's Alertbox, July 11, 1999:

Web Research: Believe the Data

By now, we know a good deal about users' behavior on the Web. For example, they demand fast download and are extremely impatient and want immediate support for their own goals. Even so, most websites are slow, internally-driven, and do not focus on solving the users' problems.

Do not ignore research: it can improve your site by several hundred percent. Even when a specific project does not have data from its exact customers using its exact site, it should use general data about the average behavior of Web users with common sites. Well, maybe your site is so unique and different that people will behave differently when they visit your site than when they visit other sites. But I wouldn't bet the farm on it. If you have data showing that you are different, then OK, but in the absence of specific data, it's best to follow the general rules.

Example 1: Delay Customer Registration

Customer registration forms are the bane of attracting new customers. It's like posting an armed guard at the door to a department store and only letting people in the store after they show two forms of ID and suffer nosy questions about their family tree.

Web users often turn away rather than having to register. Quite simply, it's not worth peoples' time to answer all your questions. Hopefully, the usability of customer registration will improve with the new Microsoft Passport initiative. But even if we get a Web-wide standard for one-click registration, users will resent being asked to register. Every click is a burden for busy Web users, but more important, users don't like parting with their personal data before they have developed a sense of trust in the site.

Customer registration (and anything else that gets in the way of allowing users to focus on what they want to do) should be minimized and made as easy and fast as possible; make sure that all such forms are usability tested.
Comply with the new Passport initiative and any other widely used mechanisms for allowing users to enter their data once and reuse it across the Web. Single log-in is a known usability requirement for all other systems, and we need it on the Web as well.
Postpone registration as far as possible into the usage process: if you ask too early before you have established your value to a new customer, you will simply turn away the prospect.

It is common for sites that violate these rules to discover that 1/3 or more of their "customer" registrations are bogus. Unless you believe that Donald Duck is a very frequent user of these sites.

Qualitative studies have long shown that customer registration hurts usability and makes users turn away. But how much does it hurt?

At the annual conference of the Usability Professionals' Association earlier this month, Marie Tahir from Intuit described the usability process used in developing several versions of QuickenMortgage.com (a mortgage site). Too-early customer registration requirement posed a major problem in the earlier version of the site. After this finding, the site was redesigned to allow users to enter valuable areas of the site without having to register. Registration was postponed to a later stage where it was truly necessary to know the user's personal data in order to provide a mortgage. As a result, usage doubled.

Two lessons:

Yes, usability works. When you fix a part of your site that users can't or won't use, you will get immediate results.
The magnitude of usability improvements is usually large. This is not a matter of increasing use by a few percent. It is common for usability efforts to result in a hundred percent or more increase in traffic or sales.

Example 2: Effectiveness of Web Marketing

In a recent study reported by Internet World and Iconocast, Forrester surveyed Web marketing executives to find out how much they used various means of promoting their site and how effective these methods had been. The results show a negative correlation between the effectiveness and use of Web marketing methods (see figure). In other words, the less results you get from a Web marketing method, the more it is used.

Scatter plot of Web marketing methods rated for effectiveness and frequency of use
Survey of Web marketing executives by Forrester Research, April 1999.

I have said since 1997 that advertising doesn't work on the Web. My original analysis was based on qualitative studies of user behavior, but there is now much quantitative data to support the conclusion:

eye-tracking studies find that users never even see the ads
click-through rates dropping from 2% to 0.5% in a few years
sales data from many sites showing that they usually don't sell a lot to those few users who do click through - paying customers usually arrive in other ways

The dropping click-through rate may be the single-most striking set of data in all Web research because the trend line is so clear and has been so consistent over the last three years. Continue the trend line out a few years (it will hit 0.1% by the end of 2000) and the conclusions are clear:

Don't build Internet business models that rely on sites getting substantial advertising revenues - except for huge sites: Yahoo can survive its current monetizing quotient (MQ) of 0.4 cents per page view because they have 310 million page views per day; you can't.
Don't make online advertising the center of the marketing plan for your own site - instead combine offline advertising with Internet-appropriate marketing methods like affiliate programs and email (to customers who ask to be notified; never send spam if you want a reputation as a reliable and high-trust site).

Many marketers are in denial of the plain-spoken message communicated by the data. They would like advertising to succeed on the Web just like it was successful in the old media they know how to deal with. So they keep running banners even though they don't work. I say, let's believe the data and move on.

I recently asked the marketing director for a major e-commerce site why they didn't have an affiliates program, even though this way of encouraging inbound links is known to be one of the best marketing methods on the Internet (the top scorer in the Forrester study). The answer was basically that the software effort to develop an affiliates program was too burdensome. Many others probably have similar reactions, which could explain why the best marketing method is the least used. This is a major opportunity for somebody to develop an utterly simple way for websites to set up affiliates programs. But even with current technology, the cost to have a geek or two handle the software side of the marketing program is much less than most spending on less-effective methods.

More Research Needed

Unfortunately, there is still too little publicly available research on Web usability. Some proprietary research exists and I am gratified by the growth in the number of companies that are running their own studies. But we also need published results since many of the findings do generalize. For example, several studies have shown that the Back button is the second-most used feature on the Web (after clicking on links), and this finding is a great counter-argument against the people who want to open new browser windows because "users won't find their way back otherwise." Oh, yes, they will.

I particularly encourage universities to conduct more research into Web use. Soon, it will be too late to study the transition from the pre-Web to the post-Web world. The control group of non-wired people will vanish. Hurry.

Previous: June 27, 1999: Content integration
Next: July 25, 1999: Metcalfe's Law in reverse

See Also: List of other Alertbox columns