What actuallymoved the needle.
A/B tests and live experiments from a real store — the numbers, the caveats, and what we'd do differently. Misses included.
A/B tests and live experiments from a real store — the numbers, the caveats, and what we'd do differently. Misses included.
From the store Ankka was built on. Where a result is a before/after rather than a randomized split, the write-up says so.
+2.6pp
Checkout-to-purchase up 2.6 points — and the entire lift arrived through one payment method.
Win — with a twist — read the test →4% vs 25%
Quick-filter pills reached 4% adoption against a 25% target — but the few who used them converted about twice as well.
Miss — thesis confirmed — read the test →≈ break-even
Week one of Google-set prices: roughly break-even in aggregate — hiding big per-product winners and losers.
Neutral overall, wild per product — read the test →−10pp
A banner redesign dropped consent acceptance by about ten points — and every consent-gated metric moved with it.
Regression, caught — read the test →Every test ran live on the Nordic baby-gear store Ankka was built on — real revenue at stake, not a lab.
A 4%-adoption feature and a −10pp banner teach more than ten cherry-picked wins. Caveats are part of every write-up.
Results feed the knowledge base Ankka’s agents draw on — to propose the next experiment on your store, and to read its result honestly.
This library is a slice of the knowledge base Ankka maintains for a store: what was tried, what it did, and what holds up. The free tracking audit is where yours starts.