Refund rules that actually work: what to grant, what to contest

This playbook is about the decision itself: once your server can answer Apple's consumption requests and Google's chargeback reviews, what should it actually say? The answer is a policy, not a reflex. Grant everything and you train refund farmers; contest everything and you burn genuine customers for pocket change. The middle path is a short ladder of rules that treats different refunds differently.

The instinct after your first refund wave is to contest everything. Resist it. The stores see your responses in aggregate, and customers you fight over $2.99 leave one-star reviews. The winning strategy is a rules ladder.

First, measure what refunds cost you

You cannot tune what you have not measured, and both stores hide the number in plain sight. In App Store Connect, open Trends, switch the measure to Proceeds, and filter Transaction Type to Refund: that line is your monthly leak. On Google Play, the financial reports in Play Console break out refunds and chargebacks per order. Write the monthly figure down before you touch a single rule; it is your baseline for judging whether any of this works.

While you are there, look at the shape, not just the size: which products get refunded, how long after purchase, and how often the same buyers come back. Those three dimensions are exactly what the ladder below keys on.

Grant fast where goodwill is cheap

Small purchase, zero usage, first refund ever? Grant it. It costs you a couple of dollars, the customer remembers the experience, and your contested-refund credibility improves for the cases that matter. A rule like price < $5 AND usage == 0 AND prior_refunds == 0 → grant handles this automatically.

Contest with evidence where consumption is real

The unsatisfied-after-heavy-usage pattern is where revenue leaks. Six days of daily usage followed by 'unsatisfied with purchase' deserves a decline preference backed by the usage curve. This is exactly the case consumption data exists for — and it's where automated evidence shines.

Always contest repeat refunders

Refund farming is real: the same account cycling purchase → consume → refund. Track lifetime refunds per user across your apps and harden the response at two or more. The stores weigh this history too; your evidence makes it visible.

Tune the ladder, don't set and forget

Run the rules for a month, then put two numbers side by side: your save rate and your review score. If the save rate climbed and reviews held, tighten another notch, for example by lowering the goodwill price ceiling. If reviews dipped, your contest threshold is catching real customers; loosen it. And remember the consent prerequisite on iOS: Apple expects the customer to have agreed, through your terms, that usage data may be shared before you send any.

Let the deadline never decide

Whatever your rules, the silent failure mode is missing the response window entirely. Make sure something — a system, not a person — is guaranteed to answer before Apple's 12 hours or Google's 24 run out.