28 January 2026 · Experiments

Designing SKU tests that survive store review

A price ladder is not an A/B test until both SKUs exist in the same countries at the same time. Stores do not owe you that courtesy.

In-app purchase analytics teams love to “test $4.99 against $6.99.” Then App Review takes nineteen days in one storefront and three in another. Play Console publishes the mid-tier overnight in Indonesia and sits in Thailand. The experiment is already contaminated, but the dashboard still has two colourful bars, so the meeting proceeds.

Write the kill criteria before you submit

SKU Experiment Studio opens with a one-page isolation sheet. You name the single difference you intend: price, trial length, or introductory eligibility — not all three. You also name the conditions that kill the read: staggered approval beyond N days, a country mix shift above a stated band, a paywall copy change, a competing campaign. If any kill criterion trips, you pause. You do not “learn directionally.” Directional learning from a dirty ladder is how catalogs accumulate folklore.

Currency rounding is a treatment

A 20% discount that lands on a clean ฿79 in Thailand and an awkward $4.79 elsewhere is not the same offer. Rounding rules are part of the treatment. Either you lock the experiment to storefronts where rounding behaves, or you accept that you are testing local price psychology, not a global percentage. In-app purchase analytics that averages them into one lift is performing science.

Do not change the paywall copy in the same week

This sounds obvious and still happens because design and monetization sit in different stand-ups. If the sheet, the headline, and the SKU all move, you have a launch, not a test. Ship copy first, wait a full billing cycle in your smallest region, then open the ladder. Patience is cheaper than a wrong annual price that takes a year to unwind.

What “survive store review” actually means

It means both SKUs are live, visible, and purchasable for the same audience definition before you start the clock. Screenshot the console. Keep the timestamps. If a reviewer rejects one identifier, the experiment has not started. The homework we mark in studio asks for those screenshots. Teams that cannot produce them are not ready to claim a winner.

If you want the isolation sheet without the six-week studio, write us via contact and ask for Signal Desk. The flagship Lab only skims this topic; the dedicated desk is SKU Experiment Studio.