NN/g eyetracking proved flat design's weak click cues cost users 22% more time

By 2017 flat design dominated the web, and with it came interfaces stripped of the cues that say what is clickable: blue underlined links, beveled buttons and drop shadows gave way to ghost buttons and text indistinguishable from body copy. The industry argued about the trade-off by taste — designers defended the clean look, usability specialists warned anecdotally that users were getting lost, and neither side had data.

Nielsen Norman Group ran a controlled eyetracking experiment to replace opinion with measurement. It took nine live pages from six domains — ecommerce, nonprofit, hotel, travel, technology and finance — and made two versions of each that differed only in signifier strength. Seventy-one general web users, each assigned to one version, were asked to find where they would click while eye-tracking equipment recorded their fixations and time.

The weak-signifier pages cost real effort: users spent 22% more time and made 25% more fixations before finding the target, and six of the nine page pairs showed different gaze patterns — attention spread across the page because nothing looked confidently clickable. On the fine-jewelry page, 86% of users on the strong version moved from the section heading to the "Shop Pearl" link versus 50% on the weak version, where nine of 24 users tried clicking the heading itself.

Just as valuable was where flat design worked: three page pairs showed no difference, and they shared low information density, traditional layouts and high-contrast targets. The study did not condemn flat design; it specified when it is safe. Its publication generated enough debate that NN/g published responses to the criticisms, and the findings fed the firm's flat-design guidance for years afterward.

Settle the taste war with measurement: matched page pairs differing only in clickability cues, eye-tracked on find-the-button tasks

Matched pairs held every design variable except signifier strength constant, so a live-page study could claim causation rather than correlation.

Fixations and task time translated an aesthetic debate into effort costs that any stakeholder could compare.

A between-subjects design with 71 randomly assigned users gave the differences statistical significance (p < 0.05).

By publishing the conditions where flat pages did no harm, the study gave designers an actionable answer instead of a verdict to ignore.

Public heatmaps made the effect visible at a glance: weak signifiers spread attention across the page, the signature of users hunting for a button.

When an industry argues by taste, isolate the contested variable in matched pairs and publish the numbers; the debate becomes a fixable problem with stated conditions.

The study became one of NN/g's most-debated pieces: the day after publication it drew 354 points and 171 comments on Hacker News, and NN/g's own account noted the article generated "a significant amount of debate about flat design," which it answered with a published response to criticisms. The findings fed a broader research program — follow-up articles on when flat design can work, how long-term exposure slowly decreases user efficiency, and what a corrected flat aesthetic would need — turning one experiment into a durable reference point for designers choosing visual styles.