Eye-tracking research that turned a clunky mobile experience into a clear path to global literature.
Words Without Borders (WWB) is a digital magazine that amplifies global literary voices, publishing translated works from diverse authors and cultures. Its mobile site is the gateway for readers to discover international literature, both renowned and emerging.
Ran eye-tracking sessions, synthesized behavioral and analytics data, and turned the findings into a research-backed mobile redesign.
The redesign cut bounce rate, sped up navigation, and surfaced multilingual content users couldn't find before.
206.7K of WWB's 417.2K users visited on mobile. Every engagement metric trailed desktop.
96% new users
Below the 68 industry benchmark. Desktop scored 78.
Mobile bounce vs the 51.6% category benchmark.
Mobile sessions vs 2:34 combined across devices.
Mobile engagement vs 46% combined. Jul 2023 – Mar 2024.
Four friction points surfaced across every mobile session.
Scattered gaze
Eye-tracking showed scattered gaze — no dominant entry point.
60% bounce
60% bounce rate. Only 13.8% reached the bottom.
Only 37% could locate it. Most misread the tag.
Links looked like text. Just 25% knew what was tappable.
Four research-driven moves — navigation, homepage, multilingual, article pages.
Clear sections, trimmed metadata, 3–4 cards per screen.
One unified filter. Predictive search. Learn as its own tab.
Explicit language labels — French, Chinese — on every card.
Underlined links, spaced credits, sort-by filters on genre pages.
Aligning on goals, defining the questions, and choosing a mixed-method approach to study mobile behavior.
Three business goals from WWB's Digital Director, translated into three questions that drove every session.
Improve nav & search discoverability — effortless mobile-first browsing to any piece of content.
Where does mobile attention actually land, and what pulls it away from the navigation?
Grow multilingual engagement — surface translated works and editorial features hidden in the IA.
Where does search break? Do "Multilingual" and "Learn" read the way we think they do?
Cut bounce, deepen scroll — tighten the homepage so readers stay long enough to find something worth reading.
Why do readers bail early? What moment would have kept them scrolling?
A mixed-method approach — behavioral data + direct user feedback.
Baselines for bounce, scroll depth, session duration — plus heatmaps and scroll maps for attention and drop-off.
Moderated eye-tracking with task-based exploration. SUS and open-ended feedback close every session.
Where users actually looked, tapped, and gave up.
Eight participants ran task-based sessions at the Pratt Usability Lab — where gaze, hesitation, and SUS scores set the baseline.
8 participants · 6 new, 2 returning · task-based exploration with SUS and open-ended feedback closing every session at the Pratt Usability Lab.
Baseline metricsFour patterns repeated across every session.
2/8 reached the bottom. Drop-off after 1–2 folds.
3/8 located it. Most read the tag as a translator.
No category guidance. Users hesitated on what they were searching.
Links looked like text. Eye-tracking caught hesitation.
Cross-cutting the lab observations against GA4 and Hotjar to separate one-off frustration from systemic friction.
GA4 + Hotjar told the same story: mobile readers drop early and bounce often.
96% new users
Browse by Topic wins
Only 20% go past fold 2
Four insights showed up in both the eye-tracking sessions and the analytics.
Gaze trailed off. 13.8% reached the bottom.
Misread in the lab. Near-zero clicks in Hotjar.
Clicks landed. Session time collapsed at layer two.
Gaze hesitated. Hotjar showed almost no clicks.
Five design moves to surface content, cut friction, and turn passive scrolling into intentional reading.
62% couldn't locate multilingual articles — most expected the tag to translate the page.
Users interpreted the label as a translation feature. Tag placement was buried inside filters and missing from preview cards.
Swap “Multilingual” for language pairs like “Chinese / English.” Split “Multimedia” into Audio and Video.
Explicit language tag
Compact card variant
75% struggled with search — filters read as post-input, and “Learn” was mistaken for a search feature.
No predictive suggestions, no category guidance, no clear separation between “search” and “learn.”
Predictive suggestions on focus. Three filters collapse to one. “Learn” becomes its own nav tab.
40% reported information overload — dense cards, long descriptions, and mixed alignment broke the scan.
No dividers. Topic showed one genre at a time. Mixed alignment. Only 13.8% reached the end.
Section headers, trimmed metadata, 3–4 cards per screen. All genres visible in Topic. Left-aligned throughout.
Only 2 of 8 could identify clickable text — author names read as plain type, spaced too tightly.
No underline, no contrast shift, no padding between adjacent names. Users missed navigation opportunities and tapped the wrong target.
Underlines and contrast on every clickable text element. Padding between author and translator names for touch accuracy.
No way to sort long article lists — the experience became overwhelm by default.
Genre pages surfaced dozens of articles with no ordering — no Latest, no Most Read, no curation.
Would be nice to see top picks or “frequent flyers” among WWB authors highlighted. — User 02
A single filter with Latest, Most Read, and curated picks — scan instead of scroll.
Clearer navigation. Surfaced multilingual content. Article pages that finally read like article pages.
Labels that feel obvious to designers can confuse real users — especially on a global platform. Eye-tracking surfaced hesitation no analytics report could show.
Clear language leads to confident navigation. Small tweaks — repositioning a tag, simplifying a filter — move the metric.
Good design isn't just what looks clean. It's what feels intuitive to real people.
