Used A/B testing to ship a winning grid layout in 14 days
+18% AOV · 95% confidence
AOV+0%£64.20 → £75.80Confidence0%Welch z-test on 11.3k sessionsTime to ship0dHypothesis to site-wide rollout
The challenge
Lumen suspected the static homepage UGC wall was tired but had no clean way to prove it. Their previous tool didn’t do attribution or sequential testing, every change was a guess.
What we did
1Hypothesis: a 4-up product-tagged grid would beat the static wall on AOV by promoting "and-also" products.
2Sequential MAB across 100% of homepage sessions for 14 days, 60% / 40% split, auto-promoting once the upper credible bound stabilised.
3Wired to the order webhook so revenue per session was attributed correctly to the variant.
“It’s the first time we’ve shipped a UGC change with the receipts to defend it in a board meeting.”
Priya Bansal · Head of Growth, Lumen Skincare
Stack
Shopify
Klaviyo
GA4
Want a result like this?
Pick your platform, we’ll show you the same numbers on your own data.
We use essential cookies to run this site and optional analytics cookies to understand how it’s used. You can change your choice anytime in our privacy policy.
All case studies
Case studies
