What UA teams should measure inside the playable

A playable is a small product surface, one that takes input and gives feedback before it resolves, yet many teams still measure it like a static banner: an impression, a click, nothing else. Treating it as a black box leaves the most useful creative evidence on the table, because the interaction itself is the part of the format that a video or static image can't produce.

Measuring inside the playable: instrument decisions, not every animation

A practical event model captures six things: first input, meaningful choice, the success-or-failure flag, replay, CTA exposure, clickout. These events reveal whether people understood the interaction and whether desire increased through play; a start and a click can't distinguish either on their own. It's tempting to instrument everything a build does, every animation frame, every micro-interaction, but that produces noise a small team can't review in time to act on it. The discipline is choosing the handful of events that map to an actual decision a person made, and skipping the ones that only describe motion.

Connect behavior to the next session

The richest analysis pairs in-play behavior with acquisition source and early product outcomes. That makes it possible to compare which experience prepared the strongest cohort once installed, rather than simply who clicked. A playable that produces a high completion rate but a weak first session is telling the team something different than one with a modest completion rate and a strong first session, and only a team pairing the two data sets will see the difference at all rather than crediting both builds equally on click-through alone.

This pairing requires the playable's event data and the product's onboarding data to share a common identifier. That's a data engineering task more than a creative one, but it's worth treating as a prerequisite rather than a nice-to-have. Without it, a team can describe what happened inside the playable and what happened inside the game as two separate stories, and never actually test whether the first caused the second.

A minimum viable event set

Teams starting from nothing don't need a sophisticated analytics stack to begin. First input, the moment a viewer stops watching and starts touching, is the single most informative event to add first, because the gap between impression and first input is usually where the largest share of viewers is lost. Meaningful choice comes next, followed by a clear success or failure flag, and only then replay and CTA exposure. Building the event model in that order gets a team most of the diagnostic value within the first two or three builds, rather than waiting for a complete instrumentation project to ship before any of it is useful.

Avoid the vanity metric trap

Completion rate is the easiest number to report and the easiest to over-trust, because a team can tune a build to make finishing easy without making the underlying motivation any more real. A playable can post a high completion rate by being simple to the point of trivial, recruiting installs that then bounce immediately once the actual game asks more of them. Comparing completion against downstream retention, rather than only against other playables' completion rates, is the check that keeps this metric honest.

The same caution applies to replay rate. Teams often celebrate it as a sign of strong engagement, but it can equally indicate a build that failed to communicate a clear resolution the first time through. A high replay rate paired with a low CTA exposure rate usually means players don't know how to finish, not that they're delighted enough to try again, and the two explanations call for opposite fixes.

Once the pairing exists, the useful question shifts from "which build performed" to "which build recruited the cohort we actually want," and those two questions produce different rankings often enough that skipping the pairing quietly costs a studio real budget over a year of campaigns.

The UA Ledger view

Instrumentation should answer a decision the team can act on. If an event can't change a brief or a build, and can't move a budget, it's probably noise. A team that resists the urge to track everything will usually end up understanding more about what actually persuaded a player to keep going, not less.

Related archive reading

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

• Rewarded Video's Retention Mandate, Beyond the Install
• Localisation Inside the Playable Ad: Text, Gesture, Pacing