tiktok playable ads: the build spec, the 2-second rule, and iterating like a static.

playables went industrial — 50–60k shipped a month. the edge isn't having one anymore, it's shipping one the network can read and iterating it faster than it fatigues. the operator build spec and iteration loop.

in october 2025 there were 50,000–60,000 new playable ads shipped every single month — and 56% of the top-30 mobile game creatives by spend were playables, up from 30% a year earlier (Sett / AppMagic). playable production has gone industrial. which means "do you have a playable" stopped being the question a while ago. the question now is whether you can ship one the network can actually read, and iterate it faster than it fatigues. this is the operator version — the build spec, the one rule that decides whether your playable gets a fair test, and how to iterate it like you'd iterate a static.

a tiktok playable is not an applovin playable

start here, because most guides skip it. playables are the highest-converting format there is — Liftoff's 2025 index puts them at 8x the impression-to-install rate of non-playables for top spenders, and 16x for everyone else. but that firepower concentrates where playables run at volume: applovin and unity. tiktok and meta are video-first surfaces. so the practical read for a lean team is simple — put your heavy playable volume into applovin/unity, and treat the tiktok playable as a tighter, hook-led unit built to tiktok's own rules. do not just re-zip your applovin asset and hope. different surface, different spec, different opening. (new to the format? start with the anatomy of a playable, then come back here for the tiktok specifics.)

the build spec, exactly

most "failed" playables are spec fails, not creative fails — the network throttles or rejects them before the idea ever gets tested. tape this to your monitor before you brief anything (per current build guides — Segwise; always re-check tiktok's live ad spec, these drift):

• keep the interactive loop around 10–25 seconds (30s hard max)
• deliver as a .zip under 5MB, containing an index.html and a config.json
• auto-transition into the playable after roughly 10 seconds of inactivity
• one playable per ad group — you can't stack them

none of that is creative judgment. it's a checklist. get it wrong and the best concept in the world never leaves the launchpad.

the first two seconds are the only part that gets a fair test

show the interaction cue — the tap, the swipe, the thing the player actually does — inside the first 2 seconds. on an algorithmic feed, a weak opener gets gated before it ever gathers enough data to prove the rest of your loop. the opening interaction is your hook, and it carries the whole unit exactly like a static's first frame does. so when a playable underperforms, don't rebuild the playable — build three or four new openers on the same proven body and let the market pick.

iterate a playable like you iterate a static

this is the reframe that changes a team's output. stop thinking "make a better playable." start thinking "run more variants of a working one." it's the same loop you already run on statics:

• concept — the core interaction and angle. the expensive part; you make few of these.
• iterate — on a proven body, swap one element at a time: the opener, the interaction cue, the reward beat, the end-card CTA. cheap, fast, and where most of your wins actually come from.
• read at the element level — not "playable A beat playable B," but "the tap-opener beat the swipe-opener." that's a learning you can carry to the next title.

vary the opener first, always — it's the highest-leverage element on a video-first surface. only rebuild the whole asset when the core interaction itself is dead, not when the hook simply fatigued.

cadence and kill rules

the studios winning on these networks aren't the ones with better ideas — they're the ones testing more. a workable rhythm for a lean team, per priority title (directional — tune it to your budget): roughly 2–3 new concepts and 6–10 variants a week. give each variant enough installs to judge before you call it, and remember the failure mode on a video-first surface — the playable fatigues on its opener first. so your refresh is usually a new hook on a live body, not a whole new playable. (the leading signals for catching that early are the same ones in our creative fatigue playbook.)

one non-negotiable: it has to be the real game

a playable that teaches a mechanic your game doesn't have buys you installs that churn on day one. the honesty rule that governs ugc governs playables even harder, because the player literally just played a version of your game. exaggerated-but-true is the line — cross it and D1 retention sends the bill.

where this leaves a lean team

playable volume is table stakes now; 50–60k a month says so. the win isn't one great playable — it's throughput of readable, spec-clean variants pointed at the right channels. that throughput — generate on-brand statics and playables, read what's converting, iterate the winners — is the exact loop we built numu to run for game studios. tell us what you're launching, and we'll send you our tiktok-playable build and QA checklist.