Why Retention Kills More Prototypes Than CPI Does

A strong CPI means people are interested enough to install. Weak retention means they are not interested enough to stay. Publishers pass on more prototypes for retention than for install cost, because a cheap install that does not stick is a loss at any scale. The prototype that clears CPI and fails retention is the most common outcome, and it is the most fixable.

Why this matters

Teams optimise for CPI because it is the first test and the most visible number. A good CPI feels like validation, and it is, but only of the creative and the concept surface.

What happens after the install is a different question entirely. A player who installs and leaves within a minute cost money to acquire and returned nothing. At scale, that arithmetic is what kills games.

Publishers learned this from experience. The games that clear CPI but fail retention are far more common than the reverse, and far more expensive when they reach market.

What publishers actually check

The sequence matters, and it is not just CPI.

1. CPI. Does the creative generate interest at a viable cost? This is the first gate and the most lenient.
2. Day one retention. Of the people who installed, how many came back the next day? This is where most prototypes fail.
3. Day seven retention. Of those who returned on day one, how many are still playing a week later? This separates concepts from games.
4. Session length and frequency. How long do players stay, and how often do they return? This indicates whether the loop has depth.
5. Playtime per day. The total engagement per user per day, combining sessions and frequency.

CPI tells you whether people want to try your game. Retention tells you whether they want to keep it.

Each bar differs by publisher, but the pattern is consistent: CPI is the entry exam, and retention is the coursework. The specific thresholds are covered in How Publisher Submission Bars Differ.

Why CPI passes and retention fails

Three common reasons, and each has a different fix.

The creative oversells the game. The ad shows something more exciting than the actual first session delivers. Players install expecting one experience and receive another. The mismatch between creative and game is the most common source of good CPI with poor retention.

The first session is weak. The game may be good, but the opening is not. A slow tutorial, a confusing first level, or too long before the first reward. The problem is pacing rather than design, and it is fixable without changing the core, per Why Your Tutorial Loses Players in the First Minute.

The loop does not have a return trigger. The session ends and there is no specific reason to come back. No unfinished business, no near goal, no expected reward. The mechanics that create return triggers are in Three Retention Mechanics Worth Stealing.

Reading a weak day one

Day one retention below 30% in casual and below 25% in hyper-casual is a problem. Here is how to read it.

The fix column matters. Below 15% is rarely a pacing problem. It usually means the concept is not holding, and the right response is to stop rather than optimise, per When to Kill a Game and What to Salvage.

Between 15% and 35% is where most work happens, and where targeted changes have the highest return. The opening minutes are almost always the leverage point.

Fixing day one without rebuilding

Four changes that move day one retention without touching the core mechanic.

1. Shorten the time to the first meaningful choice. If the player does not make a real decision in the first thirty seconds, they are watching rather than playing.
2. Match the creative to the actual first minute. Show in the ad what the player will experience immediately. Mismatched expectations are the cheapest retention problem to fix.
3. Add a near goal visible from the start. A meter, a progress bar, a locked reward. Something the player is close to and can see.
4. Remove interruptions from the first session. No permission requests, no full-screen popups, no ads before the first loop completion.

These are small changes with large effects. Individually, each can move day one by two to five points. Together they can double a weak number.

What we would do

Run the CPI test and the retention test at the same time whenever the publisher allows it. Waiting for CPI results before measuring retention adds a cycle, and the retention data is more important.

If day one is weak, watch session recordings from the first minute rather than guessing. The drop-off point is usually visible in the first five sessions, and it is usually earlier than expected.

Then make one change at a time and retest. Stacking multiple fixes makes it impossible to know which one worked, and the one that worked is the one worth keeping.

The short version

• CPI measures interest. Retention measures whether the game delivers on that interest.
• Publishers pass on more prototypes for retention than for install cost.
• The most common failure is a creative that oversells the first session.
• Day one below 15% usually means pivot. Between 15% and 35% is fixable with targeted changes.
• Shorten time to first choice, match the creative, add a near goal, and remove first-session interruptions.
• Test CPI and retention together when possible, and fix one thing at a time.

Check your day one number against these ranges today. If you want help diagnosing a retention problem in a prototype that passed CPI, tell us about the build.

Related reading: How Publisher Submission Bars Differ, Why Retention Beats Downloads, and Three Retention Mechanics Worth Stealing.