Nobody watches a playable for a set duration. They arrive, decide within about a second whether to stay, and then leave at whichever moment stops rewarding them. Total length is a consequence of that, not a target, which is why creatives built to a stopwatch tend to be badly paced at one particular point and fine everywhere else.
The useful unit is the beat. There are five, they always occur in the same order, and each has a job that can be checked on its own.
The five beats
Hook. Something is already happening when the creative appears. The job is to prevent the dismissal, and it is over in about a second, not because a second is a rule, but because the decision to dismiss is made in roughly that time and nothing added afterwards reaches a viewer who has already gone.
Cue. The viewer learns what to do. A hand, a highlight, a piece pulsing, not a sentence. This should cost about a second and it is the beat most often padded, usually by an instruction screen nobody reads that delays the first interaction past the point where interest survives.
Loop. The viewer does the thing, often enough to understand it. This is the only beat whose length really varies, and in practice it runs somewhere between three and eight seconds depending on the mechanic: a merge needs more repetitions to make its point than a pull-the-pin does.
Resolution. The board is solved, the level completes, the reward pays out. It has to be unmistakably a resolution and it has to be quick, around a second. A celebration that outlasts the achievement reads as filler.
CTA. The offer, on screen, tappable, unambiguous. It should appear immediately after the resolution while the good feeling is still there, and once shown it should stay: a CTA that animates in and then out is the most avoidable conversion loss in the format.
A working default
About one second of hook, one of cue, three to eight of loop, one of resolution, then the CTA, which puts most well-paced playables somewhere between seven and twelve seconds to the CTA. Treat that as a starting shape to adjust from, not a specification. The number that matters is not the total; it is that no single beat is doing more than its job.
The placement changes the budget, not the structure
In a rewarded placement, attention was granted in advance and the unit usually cannot be dismissed for a fixed period. The loop can run at the longer end, and a second attempt after a failure is affordable.
In an interstitial, the viewer did not ask for this and can leave immediately. The same five beats have to fit into less, and the compression comes out of the loop (one solved board rather than two) while the hook, cue and resolution keep their length, because those are what carry a viewer who was not intending to stay.
The mistake is designing two separate creatives. Design one at interstitial pacing, which is the harder constraint, and let the loop run longer in the rewarded variant.
The three pacing failures
The first interaction arrives too late. Everything before the viewer's first tap is a cost they did not agree to pay: a logo animation, a loading bar, a story frame, an instruction screen. If the creative cannot be interacted with almost immediately, the interaction rate is being decided before the mechanic is.
The loop repeats past its point. The third identical match teaches nothing the second did not. A loop should either escalate (a bigger cascade, a harder board, a faster pace) or end. Repetition without escalation is the usual reason one playable feels long at eight seconds while another feels short at twelve.
The CTA waits for the loop to finish. Some viewers are convinced at second four and will install then if allowed to. A persistent CTA available throughout, with a prominent one at the resolution, costs nothing and captures them. Gating the offer behind completing the interaction trades installs for a tidier narrative.
A worked pacing audit
Take a match playable that reaches its CTA after fourteen seconds and feels long. Time each beat with a stopwatch while playing it the way a stranger would. Say the result is a two-second logo before anything moves, a three-second instruction card, a six-second loop of three identical matches, a two-second celebration, and then the CTA.
Laid against the five beats, every problem is now visible. The logo and the instruction card are a hook and a cue costing five seconds between them; replace them with a board already in motion and a hand pointing at the first move. The loop's third match teaches nothing: cut it, or make it the cascade that clears the board. The celebration can lose a second without losing its meaning. The same creative now reaches its CTA in around eight seconds, and nothing that carried persuasion was removed.
Those timings are an illustration, not a benchmark; the method is the point. Time the beats, name what each second is doing, and cut the seconds that are doing nothing.
Reading pacing from data
Three signals tell you which beat is wrong, and the useful property is that each points at a different one.
- A low interaction rate with normal impressions means the hook or the cue failed. The viewer arrived and did not start.
- A healthy interaction rate with a low completion rate means the loop failed: too long, too hard, or it stopped escalating. Where the drop-off clusters tells you which.
- A high completion rate with a weak click rate means the CTA beat failed: it arrived late, it was quiet, it moved, or the resolution did not feel like an achievement worth acting on.
Only the first of those is visible without in-ad instrumentation. Interaction and completion signals have to be emitted from inside the creative through whatever the network's own API supports, and what is available differs by network, which is worth settling before you rely on the numbers.
Should it loop?
If the viewer fails, a second attempt is usually worth offering, because a viewer who failed and wants to try again is engaged. If the viewer succeeds, restarting the same board is worse than leaving the CTA on screen. You have already made your claim, and replaying it dilutes the resolution you just delivered.
In a rewarded placement where the unit persists for a fixed duration, an idle state after the CTA (the scene continuing to move gently, the button pulsing) holds attention better than a frozen frame, without taking the offer away.
A pacing checklist
- Can the viewer interact within the first second or two?
- Is the cue shown rather than written?
- Does each repetition of the loop escalate, or does the loop end?
- Is the resolution unmistakable, and is it short?
- Is a CTA available throughout, and prominent at the resolution?
- Does the CTA stay on screen once it has appeared?
Create playable end cards in minutes—no code required.
Open Playable Ads Maker