When playables convert better and retain worse

Playables carry a reputation for delivering better-quality installs than video, on the reasoning that a person who has tried the game knows what they are getting. Sensor Tower's State of Mobile 2026, published in January, put playables at around 13% of creative share, nearly doubled year on year, and part of that growth rests on the quality argument. The argument is often right. It's also often wrong in a specific and predictable way, and when it's wrong the dashboards don't tell you why.

The thesis: a playable that converts better and retains worse is not a bad playable. It is a playable whose mechanic promised a different game from the one that gets installed. The failure sits in mechanic selection rather than in the format, and you can diagnose it from the build before a single impression runs.

The mechanism: a playable is a promise with a demonstration attached

A video ad makes a promise the viewer has to take on trust. A playable lets the viewer test the promise for fifteen to thirty seconds. That is why it converts: the person who installs after playing has already had one satisfying loop and expects more of the same.

The retention problem arrives when "more of the same" isn't what the game delivers. Three common mismatches produce a high-convert, low-retain playable, and the first one is the easiest to spot: the playable isolates the most satisfying thirty seconds of a game whose real loop is slower, so a merge title's playable shows rapid merges while the game itself front-loads tutorial and resource gating.

• The playable uses a simplified version of the core mechanic that removes the friction the game depends on for depth. The install cohort liked the simple version and finds the real one harder.
• The playable is genre-adjacent: a puzzle-flavoured mini-game bolted onto a strategy title because puzzle mechanics test well. This is the pattern the desk has written about under the heading of borrowed mechanics, and it is the most reliable producer of the convert-retain gap.

In each case the playable does exactly what a playable should. It demonstrates. The demonstration is simply of a slightly different product.

Why the gap is hard to see in standard reporting

Playable and video cohorts are usually compared on install rate and CPI, where the playable wins, and on blended retention, where the difference washes out because the playable's cohort sits in with everyone else's. The convert-retain gap only shows when retention is cut by creative unit, and then only when enough of the install volume comes from a single playable to make the cut readable.

There's an incentive problem stacked on top. The playable producer answers for conversion, since that's what arrives first and what the network reports, while the retention question lands with a different team weeks later, usually framed as a product issue. The person who chose the mechanic never sees the retention curve for the cohort that mechanic produced.

This is the second-order effect most coverage misses: a good playable can make the product look worse. A team that ships a high-converting mismatched playable will see day-one retention dip across the whole title as the mismatched cohort grows, and the natural diagnosis is onboarding. Weeks can go into tutorial rework when the actual fix is a different demo mechanic.

A decision rule before the build

Mechanic selection should pass a single test before production begins: can a player who completes the playable describe the game's first real session accurately?

Operationally, that means putting the candidate mechanic beside the first ten minutes of the live game and asking three questions.

First, is the input the same? Tap versus drag versus swipe matters more than it looks. A drag-to-merge playable for a tap-to-place game produces a cohort that expected a different physical feel.

Second, is the pace within range? Take the number of satisfying moments per minute in the playable and in the first session. If the playable delivers three times as many, you've trained the cohort to expect a game that doesn't exist.

Third, is the goal the same? A playable where the goal is "clear the board" for a game whose goal is "build the base" will convert well among board-clearers and lose them at the base.

Any mechanic that fails two of the three belongs in a video concept instead, where the promise is softer and the retention penalty smaller. Or rework it until it passes.

An illustrative read of the numbers

Picture two playables for the same title, both live for three weeks at similar spend. Playable A uses the actual core loop, slightly accelerated. Playable B uses a snappier adjacent mini-game. B's install rate is meaningfully higher and its CPI lower. On a blended dashboard, B is the winner.

Cut by creative, A's day-seven retention sits at the title's baseline while B's sits noticeably below it. Work the cost per retained day-seven user rather than cost per install and the ranking flips, often by enough to change the production roadmap. B is not worthless. It is a top-of-funnel tool that should be run with a spend cap and paired with an onboarding path that meets the expectation it set, if the product team is willing to build one. A is the asset to iterate.

The number to add to the playable spec is a retention target per creative, agreed with the measurement lead before the build, and reported back to the producer at day seven and day fourteen. Once the producer sees that curve, mechanic selection improves on its own.

Where the format is not to blame at all

Sometimes a playable retains worse for reasons that have nothing to do with the mechanic. Heavy builds that only load fully on high-end devices skew the install cohort toward users who were going to churn anyway, because the device filter is also a patience filter. Aggressive end cards that fire the store link on any tap capture installs from people who were trying to keep playing. Both look like a convert-retain gap and both are QA findings, not creative findings.

Check the build for those two before rewriting the mechanic. A playable that installs people who did not mean to install will always look like a retention problem, and it is the cheapest one to fix.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Fake difficulty and real difficulty in playable design
• Holiday Creative Hooks for Mobile Game Ads That Convert