Senior Product Designer
UX · Experimentation · Mobile
Background
Shoppers arrive ready to buy, with little patience.
Roundforest runs high-volume Google campaigns that send shoppers, mostly in the US, to data-driven "Top 10" lists on BestReviews.Guide. Every purchase triggered by a click-out earns an Amazon commission, so even tiny friction costs real revenue.
Problem
High traffic, low engagement, and a ranking to protect.
Data showed engagement was too low for our Google ranking. To rank higher we needed more engagement, without harming click-out rate and conversion, the metrics that pay.
Approach
Three initiatives, each with its own hypothesis.
No big redesign. Over four months we ran three separate initiatives along the shopper's path. Each one started from a hypothesis, and live A/B tests decided what stayed.
- Initiative 1 "Narrow down" filters List page · mid-funnel Quick filters will raise engagement and speed up buying decisions. +5% engagement
- Initiative 2 Mobile product page Product page · bottom-funnel A page that continues the list, with less density, will help shoppers click out with confidence. Low reach, high purchase rate
- Initiative 3 Product cards redesign List page · product cards Richer cards with more specs will support more confident decisions. Hurt click-out
How every test ran
Observe→Hypothesize→Design & A/B test→Scale or drop
Initiative 1 of 3 · Filters
"Narrow down" chips to help shoppers decide faster.
Hypothesis
Giving shoppers quick, clickable filters will raise engagement and guide them to faster buying decisions.
Shoppers arrived with high intent but struggled to find their match in long lists. Inside this initiative we ran three rounds of A/B tests.
A/B test 1
Is there demand for filtering?
Two variants against the baseline, to test the concept and refine the copy and interaction.
Significant EPU upliftEngagement trending up, not yet significant
Demand was there, but the chips weren't visible or inviting enough.
A/B test 2
Make the chips visible, without distracting.
Heavy elements were known to pull shoppers off task. So I kept the chips lean, raised their contrast, cut extra copy, and made them sticky while scrolling after seeing users scroll past them.
+5% engagement, statistically significant
A/B test 3
Guide with a smart default.
A pre-selected filter signals right away that the chips are interactive and gives shoppers a starting point.
Engagement improved further
Initiative result
- Result: Hypothesis confirmed. Filters raised engagement by 5% and lifted earnings per user.
- Key insight: Just having filters on the page made the list feel more credible, and that made shoppers more likely to buy.
Initiative 2 of 3 · Product page
A product page that continues the list.
Hypothesis
For fast-scanning mobile shoppers, a product page that keeps the list's patterns and cuts density will reduce friction and lead to more confident click-outs.
Final click-out decisions happen here, and the purchase itself happens on Amazon. I kept the key UI patterns consistent with the list so the whole path feels like one flow.
Initiative result
- Result: Partly confirmed. Few shoppers scrolled and engaged with the page, but the ones who did went on to buy.
- Key insight: The page didn't fit quick deciders, but it converted shoppers who research before buying. It belongs to a different user intent.
Initiative 3 of 3 · Product cards
Richer product cards, tested against the baseline.
Hypothesis
Cards with more specs and interactive elements will support more confident decisions, especially for an older, less digital-native audience.
I designed the new cards with high contrast, clear borders and strong separation for readability.
Initiative result
- Result: Hypothesis rejected. Shoppers used the new elements, but the density broke the page's hierarchy and hurt click-out. Not rolled out.
- Key insight: For high-intent shoppers, clarity and speed beat feature richness.
Outcome
Small changes, multiplied by millions of sessions.
+5%
engagement, significant once the filters became sticky.
+2%
earnings per user (EPU). At millions of sessions, a major revenue win.
2–5%
uplift in conversion metrics across A/B iterations.
Learned
Optimize the signals you can move. Keep mobile simple.
Design for proxy signals. In fast e-commerce, improving upstream behavior like engagement and readability is the most reliable way to move downstream conversion.
Friction compounds at scale. Small UX fixes become significant across millions of sessions.
Mobile needs deliberate simplification. Clarity and low density beat extra features for quick scanners, as the failed experiment showed.