Genre benchmarks are a starting point, not a target

Benchmark reports are one of the more useful things published about mobile games. They tell you roughly what retention, conversion, and session length look like in your genre, which is genuinely hard to learn any other way.

They also get used as targets, and that is where they cause damage. A benchmark describes a distribution of other people's games. Optimising toward it is optimising toward the average of a set you are not in.

What a benchmark is actually averaging

A genre benchmark is computed across many titles that share a label. Within that label, the games differ enormously — different session structures, different economies, different acquisition strategies, different audiences.

The average across them is a real number and describes no particular game. A puzzle title with a heavy live-ops calendar and one with none both count as puzzle. Their retention curves have different shapes for structural reasons, and the average of the two describes neither.

This is not a criticism of benchmark reports, which are usually clear about being aggregates. It is a criticism of how they get used — as though the average were a standard your game should meet rather than a description of a varied population your game sits somewhere in.

The second thing being averaged is acquisition. Retention and conversion depend heavily on where players came from. A game acquiring through highly targeted campaigns has different numbers from one acquiring broadly, before any difference in the game itself. Benchmarks blend acquisition strategies, so the number carries a marketing assumption you may not share.

The specific ways chasing one goes wrong

The most common damage is optimising a metric by changing who you measure rather than improving anything.

If your conversion rate is below benchmark, one reliable way to raise it is to acquire fewer, better-targeted players. Conversion goes up because the denominator got smaller and more qualified. Nothing about the game improved, total revenue may have fallen, and the benchmark comparison now looks healthy.

The same works for retention. Tightening acquisition raises retention numbers immediately, because the players who would have churned were never acquired. A team measured against a benchmark has a strong incentive to do this and it is frequently the wrong business decision.

The second failure is copying structure to match numbers. A game with below-benchmark session length can raise it by adding time-consuming steps, and this reliably works. Whether players enjoy the longer sessions is a separate question that the metric does not ask. Session length is an output of a good game, not an input that produces one.

The third is chasing a number your design cannot produce. Some games are structurally short-session. Their retention curve looks different from the genre average because the genre average includes long-session games. Pushing that game toward the benchmark means making it a different game, and the honest question is whether that is what you want rather than whether the number should match.

What benchmarks are good for

Used as a sanity check rather than a target, they are valuable.

A number that is wildly outside the benchmark range is worth investigating, because it usually means either something is broken or your game is doing something genuinely unusual. Both are worth knowing. A first-day retention far below the range often means a technical problem, an onboarding failure, or an acquisition source sending unqualified traffic — all of which are real and findable.

A number inside the range tells you the thing is not obviously broken, which is a useful thing to establish before spending a quarter improving it. Plenty of optimisation effort goes into metrics that were already fine.

The relative comparison is more useful than the absolute one. If your first-day retention is at the genre average and your seventh-day is well below it, the gap between them is the finding — something in the first week is losing people at a rate the genre does not. That is actionable in a way "we are below the conversion benchmark" is not.

See how we read a game's numbers in its own context →

The comparison that beats the benchmark

Your own game last quarter is a better comparison than another studio's average, for one reason: it holds constant everything the benchmark cannot.

Same genre, same audience, same acquisition mix, same economy, same content. A change between two of your own cohorts is attributable in a way a difference from an industry average never is.

The second-best comparison is between segments of your own player base. If one market, platform, or acquisition source performs differently from another, that difference is real and the explanation is findable. Benchmarks cannot tell you anything about that, and it is usually where the actionable variance lives.

Benchmarks are best treated as the answer to "is this number plausible" and never as the answer to "is this number good". The first is a question they can answer. The second depends on your game, and the games in the benchmark are not it.

Talk to us about measuring your game against itself →