Why a five-star average tells you almost nothing

Who gets asked

Most ratings do not arrive on their own. They are collected by a prompt inside the app, and the app chooses when to show it. Good developers show it after something has gone well: a workout finished, a file converted, a level completed.

That is not cheating, but it does mean the sample is drawn from people at the moment they were happiest. Someone who deleted the app in frustration on day two was never asked. The average is therefore not the average experience. It is the average of a filtered slice of it.

Averages hide their own shape

A 4.5 built from mostly fours is a steady, unremarkable app. A 4.5 built from a wall of fives and a hard floor of ones is something else: an app that works beautifully for most people and fails completely for some. Those are opposite situations behind the same number.

This is why the distribution bar is worth more than the headline. A visible group of one-star reviews on an otherwise strong app is usually a specific, reproducible failure: a device that is not supported, a region that is not covered, a sync that breaks. Read those. They are the most informative reviews in the listing.

Ratings reset, reputations do not

On some platforms a developer can reset the rating when a major version ships. That is defensible, because a rewritten app is arguably a different app. It also means a high average can be four weeks old and describe software that no longer exists.

Check the review dates alongside the number. A 4.8 with nothing written in the past year is a different signal from a 4.3 with a steady flow of recent, specific reviews.

What actually predicts a good app

Three things do more work than the star count. The update history, because an app updated regularly is one someone is still paid to care about. The developer's replies to reviews, because the tone of those replies tells you what support will feel like. And the specificity of the complaints, because vague anger means little and a detailed bug report means a great deal.

None of that takes long. It is roughly the same two minutes people spend staring at a number that was never designed to answer their question.

A note on the ones that look too good

Very high averages on very new apps, with short, enthusiastic, oddly similar reviews, are the pattern worth being sceptical about. Review manipulation exists, platforms remove it unevenly, and the shape it leaves behind is recognisable once you have seen it.

The honest version of this advice is unglamorous: ratings are one input, they are noisier than they look, and the written reviews from the past few months are worth more than the number above them.

Appiqon is a paid membership. This website is promoted by Globway International B.V., which receives a commission from our partners when someone signs up through a link here. This page names no apps and endorses no products; it describes how to judge them for yourself.

← All guides