Why do players quit a game?
Most quits fall into five groups, and each one leaves a different trace in your data. The table maps each reason to the signal that shows it. Use it as a checklist when retention drops.
| Reason | Signal in your data | Where to look |
|---|---|---|
| Crash or bad performance | app_unclean_exit, critical errors, short last session | Game health board, errors per session by build |
| Difficulty spike | A level with a high fail rate and few completes | Levels table, level funnel |
| Confusing first hour | A large drop between two early steps, long median time | First session funnel |
| Nothing new to do | Long sessions, then no return after a content milestone | Retention by entry event, user paths |
| Finished the game | The last level or ending event, then no return | Completion funnel, achievement rates |
Sources: Manalens metrics, error tracking and user paths documentation.
When does a player count as churned?
Pick a rule and keep it. In Manalens, the User Lifecycle chart in Audience insights marks a player as Churned after more than 30 days with no activity, and At Risk after 14 or more days with no activity after earlier activity. New players were first active in the last 30 days. Resurrected players came back after a gap of 30 or more days.
These rules fit games that people play over weeks. For a short premium game, 30 days of silence can simply mean the player finished. Read the lifecycle counts next to completion before you call it churn.
Sources: Manalens audience insights and metrics documentation.
Is a player who finished the game churned?
No. A player who reaches the ending of a finite game is a success, even if they never launch it again. Separate the two by sending an event at the ending, such as game_complete, and by tracking chapter or act milestones.
Steam gives a public view too. The GetGlobalAchievementPercentagesForApp Web API returns the global achievement percentages for an app. If you have an achievement for each chapter, the drop between chapters shows where players stopped, across all players, even those who never launched the build with your SDK.
Sources: Steamworks ISteamUserStats Web API documentation.
How do I find where players stop?
Look at the last thing players did before they left. Add a User paths chart with session_end as the end step. It shows the events that lead to the end of a session, so you can see which level, menu or screen comes right before a quit. Keep the conversion window short, such as the default of 2 hours, to stay inside one session.
Then read the Levels table. It shows, per level, the players who started it, the completes and fails, and the fail rate: fails divided by completes plus fails. A level where many players start, few complete and the fail rate jumps is a wall.
Finally, compare the lifecycle of players who reached a milestone with those who did not. A retention chart with the milestone as the entry event shows whether it changes how long players stay.
Sources: Manalens user paths and metrics documentation.
How do crashes and bugs show up as churn?
As short last sessions, unclean exits and players who never return after an error. The Godot and Unity SDKs send app_unclean_exit on the next launch after a crash, in the crashed session. The Game health board shows crash-free sessions, crash-free players, errored sessions, errors per session by build and the top errors.
A player who crashes and never launches again sends no app_unclean_exit, because there is no next launch. So also check players whose last event was an error, and compare D1 by build. A drop in D1 that starts on the day of a patch, with a rise in errors per session for that build, points at the patch.
Sources: Manalens error tracking documentation.
What do Steam refunds tell me about churn?
Refunds are churn with a reason attached. Steam refunds a game bought in the last 14 days with less than two hours of playtime. In the Steam Sales and Activations Reports site, the refund data groups the reasons players gave by category, and you can expand each one to read the notes. You can switch between the recent week, month and lifetime.
Steamworks warns that the week view is volatile after launch and after discounts, because some buyers from a sale ask for refunds later. Simon Carless of GameDiscoverCo wrote in 2020 that, in his experience, refund rates vary from 4% to over 20% and average 8 to 9%. Treat that as one experienced view, not a benchmark.
Read refund notes next to your data. If many notes mention performance, check crash-free sessions and errors by build. If they mention difficulty, check the Levels table.
Sources: Steam Refunds policy; Steamworks Refunds documentation; GameDiscoverCo (12 October 2020).
Which players are at risk right now?
Players who were active before and have been silent for 14 or more days. Open Users, then Audience insights. The User Lifecycle chart counts the players who are New, Returning, At Risk, Churned or Resurrected in the date range.
A rising At Risk count after a patch or a content drop is an early warning. It shows up before the Churned count, which needs more than 30 days of silence. Compare the At Risk count by build or storefront with the App filters. If one build has many more players at risk, start your search there.
Activity means page views and custom events. Identify events do not count, so a sign in alone does not keep a player active.
Sources: Manalens audience insights and metrics documentation.
How do I test a fix for churn?
Ship the fix in a new build and compare build cohorts on the same metric. If you lowered a difficulty spike in level 4, compare the level 4 fail rate and D7 for players who started on the old and the new build. Wait for complete periods and a few hundred players per cohort.
Do not ask the data to do everything. Short playtests and player interviews explain why a step is hard. Analytics tells you how many players it affects and whether your fix worked.
Sources: Manalens retention documentation.
How do I find my churn reasons in one place?
Manalens is the analytics and growth engine built for games. Install the Unity or Godot SDK, then read the Levels table, the Game health board, user paths and lifecycle stages next to your Steam data.
Sources: Manalens product documentation.
Frequently asked questions
What is churn in a game?
Churn is the share of players who stop playing. Each tool sets its own rule. Manalens marks a player as Churned after more than 30 days with no activity and At Risk after 14 or more days. For a finite premium game, separate players who finished from players who quit.
Where do I see why players asked for a refund on Steam?
In the Steam Sales and Activations Reports site. Select a package, then open Refund Data. Steamworks groups the reasons players gave by category, and you can expand each category to read the notes. You can switch between the recent week, month and lifetime views.
What is a normal Steam refund rate?
Valve does not publish one, and Steamworks says rates vary widely between games. Simon Carless of GameDiscoverCo wrote in 2020 that, in his experience, rates vary from 4% to over 20% and average 8 to 9%. Compare your rate with your own weeks and sales.
Can analytics tell me why players quit?
It tells you where and how many, not always why. Level fail rates, user paths, crash data and funnels show the step and the size of the problem. Playtests, reviews and refund notes add the reason. Use both, then test a fix with build cohorts.
Sources and implementation
- Steamworks documentation: Refunds (Valve, read 18 November 2026) ↗
- Steam Refunds policy (Valve, read 18 November 2026) ↗
- GameDiscoverCo: Steam refund messages, the silly and the sensible (Simon Carless, 12 October 2020) ↗
- Steamworks documentation: ISteamUserStats Web API (Valve, read 18 November 2026) ↗