A rating count with no denominator tells you nothing

You open App Store Connect and see 4,000 ratings. Is that good?

The question has no answer as asked, and that is the whole problem. Four thousand ratings is remarkable for an app with a few thousand monthly installs and dismal for one with a few hundred thousand. The number is a numerator with nothing under it, and every instinct about whether to add a review prompt, chase reviews harder, or leave it alone depends on the missing half.

The obvious fix is to divide by downloads. That fix is where most of the danger is, so it is worth walking through what breaks.

Apple does document that ratings matter

Worth establishing first, because this site is normally the one saying Apple documents nothing about ranking. On ratings, it does. Apple's App Store search page states:

"Search results are based on a number of factors, including text relevance (matches for your app's title, subtitle, keywords, and primary category), as well as user behavior (downloads, ratings and reviews, and more)."

And, more directly:

"Ratings and reviews appear on your product page and in search results, and can influence how your app ranks in App Store search."

So the interest in the number is not misplaced. What Apple does not say is how much, in what direction, or against what baseline — which is exactly the gap a denominator is supposed to fill.

The first lie: the two numbers do not share a unit

Nobody outside Apple has your download figures. What is available is a third-party estimate, and the moment you divide one by the other you have combined two quantities that are not measuring the same thing:

Their quotient is not a ratings-per-install rate. It is not a percentage. It is not "x% of your users rate you". Anyone rendering it as any of those is presenting a unit that does not exist.

What it is good for is ordering. Compute the same mismatched quotient for every app in a cohort — say the 25 apps sharing your search result — and the mismatch is identical across all of them, so position in that ranking carries information even though the number itself does not. Appstro's rating benchmark leads with a percentile for exactly this reason and treats the raw index as secondary.

The second lie: a floor is not a measurement

This one is the reason the tool exists in its current form.

Download estimates for small apps often arrive as a bounded string — < 5k — rather than a number. Underneath, the placeholder value is 1000. Divide by it and you have not made a small estimate; you have invented a denominator.

Measured against live data: an app with 91,101 lifetime ratings and a < 5k estimate scored 91.1 on the index, against a cohort median of 0.53. That is roughly 170 times the median, entirely manufactured by the placeholder, and it would have sat at the top of the ranking looking like the best-rated app in the category.

So apps whose estimate is a floor are excluded and counted separately, on both sides of every comparison. Typically that is around 8 of 25 apps in a cohort — a third of the sample, removed rather than silently mis-scored. If a tool shows you a benchmark and never mentions exclusions, ask what it did with these.

The same caution applies to precision. The estimates are coarsely rounded: measured on 20 August 2026 across the live US top-100 free chart, 100 apps returned only 24 distinct download values — 7k, 20k, 100k, 500k, 3m, 20m and so on. Ninety-eight of the hundred carried a single significant figure; only 11m and 12m had a second digit. A gap of a few percent between two apps in a ranking built on those is noise, not signal.

The third lie: age

The numerator is lifetime. The denominator is one month. An app that shipped in 2013 and now gets modest downloads has been accumulating ratings for over a decade against a denominator that reflects last month. It will score beautifully without being better at asking.

This is a straightforward confounder and it is surfaced rather than buried: when the app under test and its cohort differ by more than about two years in age, the comparison carries a warning naming both ages. A six-month-old app measured against a cohort whose median age is nearly nine years is not being told it is worse at collecting ratings.

The one comparison on this page with no unit problem at all is the average star rating. Both sides are Apple's, both are per-storefront, both are lifetime. When the index and the stars disagree, trust the stars about quality and the index about volume.

What "not enough data" has to look like

A missing index must read as missing. If either side is unknown — no usable estimate, no rating count — the honest output is nothing at all, not a zero. An app scored 0 sorts to the bottom of a ranking and reads as "worst in cohort", which is a claim nobody made. The same rule runs through this codebase: absence and zero are different facts and must never render the same.

About review prompts

Nothing here will tell you to add one. Where and when an app asks is a product decision, and a prompt at the wrong moment costs you a rating and a user.

What is worth stating is the boundary, because it is an account-level risk. Apple's App Review Guidelines say, in the introduction:

"If we find that you have attempted to manipulate reviews, inflate your chart rankings with paid, incentivized, filtered, or fake feedback, or engage with third-party services to do so on your behalf, we will take steps to preserve the integrity of the App Store, which may include expelling you from the Apple Developer Program."

And guideline 3.2.2(x) is explicit that apps "must not force users to rate the app, review the app, download other apps, or other store-related actions in order to access functionality, content, or use of the app."

Buying, trading or gating ratings is not a growth tactic with a downside. It is the loss of the developer account.

What this cannot tell you

• Not a ratings-per-install rate. The units do not permit one, in either direction.
• Not whether your rating volume is causing anything. Apple says ratings can influence ranking; it does not say by how much, and a correlation on your own chart is not an attribution.
• Not the "right" number of ratings for your category. There is no published benchmark, so this compares you against the apps you actually share a result with instead.
• Nothing about review sentiment. A count is a count. What people wrote is a separate question.

The work

Run your app through the rating benchmark and read the percentile, not the index. Then check the worldwide sweep, because a rating count is per-storefront and the US figure you have been staring at is often a minority of the total.