Jakob Nielsen On Usability

From Jakob Nielsen, jakob@useit.com:

Angus Davis wrote:

...a very typical post in the Mozilla.Org User Interface discussion forum by Taylor at Hotwired: "What does everyone think of moving the Aurora buttons to the UI up top with everything else and not down the side?"

There were many responses, and discussion ensued among the active community of Netscape engineers, Net-based engineers, and people who just wanted an open mind to listen to their ideas.

The correct way to answer the question of what to do with the buttons is to perform a usability study where two groups of users are asked to perform typical browsing tasks with the two alternative user interfaces. The design that allows people to perform the fastest, with the fewest errors, and that scores best in a subjective satisfaction questionnaire wins and should be used in the shipping product. If one design scores best on some usability metrics and the other design scores best on other metrics, then one will need to either prioritize the usability dimensions or design a third alternative that combines the best aspects of each of the original designs. Usually, though, it is painfully obvious which design scores best in a usability test.

Experience from the last 20 years of software development shows that usability engineering is the best way to improve user interfaces. It is much faster to collect data about what actually works than to sit in development meetings and argue over what users might prefer. This is why all large software development organizations have their own usability groups and usability labs. Smaller software companies or fast-paced Web projects can collect great data through discount usability engineering.

A discussion among "Net-based engineers" is guaranteed to produce a horrible user interface since they are all completely unrepresentative high-end power users of the Web. Thus, their intuitions about good ways of interacting with the Web will likely be the exact opposite of an average user's needs.

Let's say that a leading Web browser will have 100 million users in 1999 and that each user will browse five Web pages per day (remember, most people are low-end users). If better button layout can shave 0.1 seconds of the interaction time for each page view, then making the right choice will save 5 million person-hours per year, adding maybe $50M to the world economy. Assume that making the design decision based on data rather than hearsay increases the probability of getting it right by 10% (actually, I think the correct number is closer to 90%-60%=30%). This means that running the usability study will be worth $5M; much more than the approximately $10K it will cost to do well or $1K it will cost to do cheaply.


Jacob Levy responds

From Jacob Levy, jyl@tcl-tk.com:

While I agree with Jakob Nielsen's main point about the benefit-cost analysis of usability testing, I'm fed up with the unsupported claim he makes that "us nerds" are somehow different -- exactly opposite, in his words -- in how we want our buttons laid out or in any other aspect. Where is Jakob's evidence? Let's run a usability test and lay this old folk tale to rest once and for all.

Let me postulate a possible conclusion from this usability study that's quite different from what Jakob assumes: the power users want to have a layout that's optimal, one that will make naive users just as happy and which will meet their needs just fine. Power users do the same things as naive users -- browse web pages -- but want to do it more efficiently so that they can save time. Therefore it's extremely valuable to ask power users.

I'm willing to test my hypothesis with a usability test -- is he?

Paul Snively responds

From Paul Snively, chewy@mcione.com:

My $.02 worth on usability testing:

While I was at Apple, which was during the time that System 7.0 went from d12 to ship, it underwent extensive usability testing with a wide variety of classes of users: first-time computer users, first-time Macintosh users, hard-core DOS users, hard-core System 6 users, etc.

It's admittedly only one data point, but unfortunately, this data point bears out Jakob Nielsen: power users and especially developers *are* inclined to make choices that leave less experienced users behind, unless they've studied human interface design pretty extensively. One side-effect of learning this at Apple was that Apple gained a few new bibles (that is, books you could reasonably expect to find in 75%+ of everyone's cubicles): Tufte's "The Visual Display of Quantitative Information" and Norman's "The Design of Everyday Things" were the two most obvious. Still later, Donald Norman would become an Apple Fellow.

Most engineers respond the way Jacob Levy did to the suggestion that they no longer can summon up a newcomer's naive intuition about the system, but really, it's the equivalent of saying that an adult no longer truly remembers what it was like to be a child: the loss of that gestalt is a subtle, gradual thing that doesn't even happen consciously, and certainly isn't a net loss (or at least it shouldn't be).

Good human interface designers remind me of Robert Heinlein's "Fair Witnesses" because, like Fair Witnesses, they have to learn how to see past their assumptions.

This page was last built on 4/26/98; 7:57:52 AM by Betty Webmaster. webmaster@scripting.com