Which star rating are users actually seeing?

Your dashboard shows one number. The stores show dozens — split by country, split by device, weighted by time, and resettable on one platform but not the other. The rating you obsess over is an aggregate no shopper ever sees, and the differences between the real numbers are levers you can pull.

One app, dozens of ratings

On Apple's App Store, the summary rating has always been computed per territory. A user browsing the German storefront sees a rating built only from German ratings; a user in the US sees a different number built from US ones. You can be a 4.8 in one market and a 3.9 in another simultaneously, and nothing that happens in one storefront moves the other.

Google Play worked the opposite way — one global number — until it announced the switch in August 2021. Since November 2021, phone users see a rating computed from their own country or region, and since early 2022 the rating is also split by form factor: tablet shoppers see a rating built from tablet users, and the same goes for Chromebooks and Wear devices. Google's stated reason is exactly the failure mode you'd guess: a bug that only affected one country was dragging an app's rating down everywhere, and tablet-experience improvements were invisible under the volume of phone ratings. The rollout was disruptive enough that Google committed to warning any developer whose rating would shift by more than 0.2 stars in a significant market, ten weeks before it happened.

The consequence for you: the blended number at the top of App Store Connect or the Play Console is a reporting convenience. No user sees it. The number that decides an install in Tokyo is your Tokyo rating, on that shopper's device class.

Google's rating has a clock in it

Until 2019, your Play rating was a lifetime average — a review from five versions and three redesigns ago counted exactly as much as one from yesterday. Google announced the change at I/O that year and rolled it out from August 2019: the displayed rating now weights recent ratings more heavily. The exact formula isn't published, but the behavior is easy to observe in either direction.

This cuts both ways. A rough launch no longer follows you forever — ship the fixes, prompt your happy users, and the displayed rating moves in weeks, not years. But a bad release also surfaces fast: a crash in production shows up in your visible rating while the lifetime average would still be sleeping on it. On Play, your rating is closer to a moving average of how the app is doing now than a verdict on its history.

Apple weights nothing — but hands you a reset button

Apple publishes no recency weighting. Your summary rating in each territory is cumulative — every rating since launch, or since the last time you reset it. That last clause is Apple's alternative to Google's clock: when you release a new version, App Store Connect offers a checkbox — "Reset rating when this version is released" — and the summary rating starts over from zero.

Three details matter before you touch it. The reset applies to all territories at once — you can't reset just the market where things went wrong. Written reviews stay visible; only the number resets. And a listing with a rating count of twelve converts worse than almost any non-embarrassing rating with thousands, which is why Apple's own guidance says to use the feature sparingly. The reset earns its risk in one situation: the bad ratings are about a problem you have genuinely fixed, and your install volume is high enough to rebuild a credible count within a few weeks. Otherwise the cure reads worse than the disease.

Most ratings never came with words

Since iOS 10.3 in 2017, users can rate an app from the in-app system prompt without writing anything, and Play's in-app review flow does the same. That changed the composition of the number: your summary rating is now mostly silent one-tap responses from people your app chose to ask, at a moment your app chose. The written reviews you read and respond to are the loud minority.

Which means the biggest input to your rating isn't your response strategy — it's prompt timing. Ask after a success moment and you harvest the satisfied majority; ask on the third launch regardless of what happened and you sample everyone, including the annoyed. We wrote up the timing question separately in when to ask for ratings — it matters more than anything else in this post.

Reading the splits like a diagnostic

Once you know the rating is many numbers, the per-market view stops being trivia and starts being a debugging tool. App Store Connect shows ratings by territory; the Play Console breaks them down by country and device. Scan them occasionally and outliers jump out. One country sitting a full star below the rest usually isn't taste — it's a broken translation, a payment method that doesn't exist there, or a bug in a locale-specific format. A tablet rating far below phone is your tablet layout telling you it ships broken, and on Play, tablet shoppers are seeing that number specifically.

The splits also change what recovery looks like. On Play, a bad rating is recoverable by design — recency weighting means sustained fixes plus well-timed prompts move the number on a scale of weeks. On iOS there is no clock to help you; the honest levers are the prompt, replying to reviews, and — rarely — the reset.

The stars get the look — the listing gets the install

In search results, your rating appears next to exactly two other things: your icon and your first screenshots. The stars decide whether a shopper looks twice; what they see next decides whether they tap Get. If you're going to earn a rating recovery, don't waste the look it buys you — ShotCanvas turns raw screens into a store-ready screenshot set in minutes, so the listing next to your stars pulls its weight.