Skip to the page

Plainly Put

One sentence about building things, taken apart

HomeProduct

Data describes the present, design proposes something else

Usage data records what people do with the product they have. It cannot record what they would do with one that does not exist yet.

You can never let your data dictate design. If you do, you end up following what people currently do and never innovating.

Aza Raskin

Aza Raskin led user experience work on Firefox at Mozilla before leaving late in 2010 to start a health company, Massive Health. His father, Jef Raskin, began the Macintosh project at Apple, and the family view of interfaces has always been that they should be designed from how people think, not only from how they currently click.

In March 2009 Douglas Bowman, then a visual designer at Google, left the company and wrote publicly about why. Among his examples was a team that could not decide between two shades of blue and tested forty-one shades in between to find the one users clicked most. Another argument was about whether a border should be three, four or five pixels wide. His complaint was not that the data was wrong. It was that data had become the only accepted reason for any decision, so nobody could propose anything a test could not settle.

Raskin’s line names the limit. Data can compare options that already exist. It cannot generate a new one, and it is biased toward the familiar, since users perform best at first with what they already know. A test run on the first touchscreen phones against keyboard phones would have favored the keyboards for months.

The line goes too far in one word, though. “Never” leaves out the areas where data is the best tool available. Sign-up forms, checkout pages and the wording of buttons are small, repeated decisions where a test beats opinion, and designers who ignore the numbers there are indulging taste at the users’ expense. Data is also the quickest way to learn that a bold new design is confusing people.

A workable division is to let design propose and data judge. The idea comes from understanding people; the test checks whether it helped them. When the two disagree, that disagreement is worth studying rather than settling by rank.