Grow your conversions ★★★★★ 4.9 · 150+ brands
Talk to Our Team →
Skip to content
4.9 on Clutch from 49 reviews150+ brands grown€500M+ ad spend managedOfficial partner ofMetaGoogleTikTokSnapchat
Talk to Our Team

Playable ads / Testing and KPIs

Playable ads testing

How to test playable ads without fooling yourself

A KPI ladder, a test design that holds up on every network, and a decision table for calling winners, killing losers and iterating the rest.

Short answer. To test playable ads, change one variable at a time against a proven control, split budget evenly, and read IPM first to shortlist. Then let D7 ROAS and retention make the final call, because a playable can win on installs and lose on revenue. Wait until each variant clears the network's learning phase and a conversion threshold you set before launch.

Reviewed 19 November 2026 by the Admiral Media creative team

The playable KPI ladder

  1. Inside the adEngagement and completionDid players start, and reach the win
  2. At the storeCTR and CVRDid the ad send real intent
  3. Per impressionIPMThe first read to shortlist
  4. After installCost per purchase, D7 ROASThe read that decides scale
Impression-to-install vs other formats
8x for top game spenders, 16x for others (Liftoff, 2025)
Time to teach the core mechanic
3 to 5 seconds (Unity)
Variants per A/B test
2 to 4 for most budgets (Unity)
Gaming spend on the top 2% of creatives
53% (AppsFlyer, 2025)

Playable ad metrics

The playable KPI ladder, from first tap to D7 ROAS

Each rung answers a different question. Read them in order, and never let a rung near the top decide a scale call on its own.

In-ad metrics tell you whether the mechanic works. Click and install metrics tell you whether the ad sends real intent to the store. Post-install metrics tell you whether those players are worth paying for. Unity's advice: use early IPM to shortlist, then let D7 and D28 ROAS plus retention settle the decision, because a creative can win on IPM and still lose on ROAS (Unity creative testing guide).

Playable ad KPI ladder: definition, what it tells you and its trap
KPIDefinitionWhat it tells youThe trap
Engagement rateShare of impressions with a first interaction. Google bills playables per engagement.Whether the hook earns a tap.Network definitions differ and are mostly not public.
Interaction rateShare of engaged players reaching each tracked step. Needs custom events.Where players drop out of the loop.Without events, a long tutorial looks like weak interest.
CompletionShare of engaged players reaching the win moment or end screen.Whether loop length and difficulty fit.Easy wins inflate it without selling the game.
CTRStore clicks divided by impressions.How hard the ad pushes to the store.A persistent button raises clicks by design. Some taps are accidental.
IPMInstalls per 1,000 impressions.The cleanest early creative read.Counts installs, not good installs.
CVRInstalls divided by store clicks.Whether the store page keeps the ad's promise.High CTR with low CVR usually means accidental taps or a mismatch.
Cost per purchaseSpend divided by first purchases.Whether the playable buys payers.Needs far more volume than IPM. Late and coarse on iOS.
D7 ROASDay 7 cohort revenue divided by spend.The scale decision.Lags by a week or more, longer on iOS.

Definitions follow Unity, Unity best practices and Google Ads Help. Engagement, interaction and completion have no shared definition: write yours down first.

There is no public playable KPI benchmark

No network or measurement firm publishes benchmark engagement, completion or IPM for playables by genre. Liftoff reports relative figures only: impression-to-install 8x non-playables for top-spending game advertisers and 16x for others (PocketGamer.biz on Liftoff's 2025 index). Benchmark against your own control, per network and per country.

Test design framework

Five rules for a playable test you can trust

Most bad playable calls come from a messy test, not a bad playable.

01One variable at a time

Change the hook, mechanic, difficulty, reward or CTA, never two at once. Unity advises testing the tutorial, then the gameplay, then the CTA (Unity, May 2022).

02Mechanic before polish

Test 2 to 4 mechanics as rough builds first. Spend on art, sound and localisation only for the winner. See what drives playable cost.

03Minimum volume

Set a conversion threshold per variant before launch. Unity's worked example uses 50 installs per variant. Purchase reads need far more.

04How long to run

Wait out the learning phase, usually a few days per Unity, and never decide within hours. D7 ROAS needs a 7-day-old cohort, plus SKAN delay on iOS.

05Network by network

Keep a separate control per network and compare creative elements, not raw IPM. Build rules are in playable ad specs by network.

How to set up a playable test on each network
NetworkTest setupWatch out for
Unity AdsCreative Testing campaign, two or more creative packs including a proven control, in your scaling country.One responsive build must work in both orientations.
Google AdsUp to 20 HTML5 ZIPs per ad group, compared in asset reporting. Installs count within 30 days of an engagement.You pay per engagement, so engagement without installs is costly.
MetaCreate the ad paused, test on a real device, then activate. Standard A/B testing.Lead-in video required. Missing CTA call means rejection.
Mintegral1 to 2 playables per batch. Report by Ad Output Type to split Playable Included from Dynamic Included.Manual review takes 1 to 2 working days.
TikTokPlayables run only in upgraded Smart+.Automated delivery makes even splits hard. Read as directional.
AppLovinRun the variant beside a proven control in the same campaign.Guidelines cover build rules, not test design.

Sources: Unity guide, Unity specs, Google Ads Help, Meta, Mintegral, Mintegral data guide, TikTok, AppLovin. TikTok and AppLovin rows are our practice.

Step by step

The playable testing process in eight steps

The loop we run with game studios.

  1. Write one hypothesis

    Name the variable and the expected effect in one line.

  2. Pick the control

    Your best playable on that network, or your best video if you have none. See playable ads vs video ads.

  3. Set thresholds

    Minimum installs per variant, learning-phase wait, deciding KPI and winning margin, before launch.

  4. Build 2 to 4 variants

    Change only the hypothesis variable. Add custom events for each loop step.

  5. QA on a real device, per network

    Unity asks for a first interaction in under 2 seconds on a mid-tier device. Package each network separately.

  6. Launch in clean conditions

    Same geos, placements and time window, even budget split.

  7. Read IPM, then confirm on value

    Shortlist on IPM, scale only once D7 ROAS or cost per purchase confirm.

  8. Scale and log

    Scale the winner, retire losers, log the learning. The winner becomes the next control.

Decision table

When to call a winner, kill a playable or iterate

Read the pattern across the ladder, not one number.

Decision table for playable ad tests
What you seeCallNext move
IPM and D7 ROAS at or above control, threshold reachedWinnerScale in steps, check ROAS holds, make it the new control.
IPM above control, D7 ROAS belowIterateThe ad attracts players the game does not keep. Bring the mechanic closer to the core loop.
Low engagementIterate the hookCut intros and long tutorials. Live mechanic and one tap target on the first frame.
Good engagement, low completionIterate the loopShorten it or lower difficulty. Find the drop-off step in custom events.
High CTR, low CVRIterate the CTACheck button placement and first-tap behaviour. Match store screenshots to the mechanic.
Below control on IPM and value after thresholdKillStop spend and log it so the idea is not rebuilt next quarter.
Gap inside normal daily varianceKeep runningExtend or add budget. Do not call it on one good day.
Former winner, IPM falling while control holdsRefreshShip a variant with a new level, reward or theme.

Framework by the Admiral Media creative team, built on Unity's creative testing guide. Set numeric thresholds per title and network before launch.

Why most playables should lose

In gaming, the top 2% of creatives take 53% of ad spend (AppsFlyer, April 2025). From Q1 2025 to Q1 2026, creatives that never passed $100 of spend grew more than six times over, while those earning $10K or more grew 1.7x (Sensor Tower, October 2026). A programme that kills most playables fast is working.

Measurement limits

SKAN, privacy and what you cannot measure on iOS

On iOS the ladder gets shorter and slower. Plan the test around what measurement can tell you.

A playable cannot measure itself: Unity bans self-fired attribution and network calls, so results come from the network and your measurement partner (Unity specs). Google's SKAdNetwork rules tie click attribution to a click that lands on the App Store (Google Authorized Buyers). For iOS users who have not opted in, data arrives through SKAdNetwork, which Apple is extending with AdAttributionKit (SplitMetrics).

SKAN 4 postback windows and what they mean for a playable test
PostbackWindowDelayMeaning for a test
Postback 10 to 2 days24 to 48 hoursEarliest iOS value signal. Too early for D7 ROAS.
Postback 23 to 7 days24 to 144 hoursThe iOS D7 read lands up to 6 days after day 7.
Postback 38 to 35 days24 to 144 hoursFor D28-style checks, too slow for weekly calls.

Source: Adjust, How SKAdNetwork 4 works. Four crowd anonymity tiers mean low-volume campaigns get coarse values (none, low, medium, high) instead of 64 fine values.

  1. Shortlist where data is richest

    Pick mechanics on Android or high-volume iOS campaigns, then confirm the winner on iOS at campaign level.

  2. Fund fewer variants well

    Thin variants fall into lower privacy tiers and return coarse data.

  3. Put payer signals in postback 1

    Map early events that predict D7 value to the first conversion window.

No study measures SKAN's effect on playables

We found no public source measuring how SKAN changes playable performance against video. Treat iOS tests as slower and coarser.

Fatigue monitoring

How to spot playable fatigue before it costs you

No network publishes a fatigue threshold, so set your own.

Claims that playables last a set number of days longer than video could not be traced to a primary report, so we do not quote them. Unity's advice is to keep iterating to avoid user fatigue (Unity), and Mintegral asks for 1 to 2 new playables per batch. Fatigue is a trend against the creative's own history: if the control falls too, the market moved; if only the playable falls, it is tired.

  1. Record the baseline

    IPM and D7 ROAS from the first stable week, per network and main country.

  2. Set the trigger before launch

    The sustained drop from baseline that starts a refresh.

  3. Check weekly against the control

    Compare both trends over the same days.

  4. Refresh, do not rebuild

    New level, reward or theme on the winning mechanic, with the next variant built in advance. See AI playable ads.

Playable ads best practices

Playable ads best practices: a 12-point testing checklist

Each item prevents a common wrong answer.

  1. Hook in 3 to 5 seconds

    Unity's close button can appear after 5 seconds.

  2. Live mechanic on frame one

    No logos or cinematic intros. One tap target, one hint.

  3. Length set by the mechanic

    Unity: 20 to 40 seconds for simple loops, over 2 minutes for puzzle, merge and mid-core.

  4. Persistent and end screen CTA

    A button visible during play earns more clicks.

  5. Real gameplay only

    The UK ASA banned four game ads on 4 November 2026 for gameplay not in the game.

  6. Fast load, real device

    First interaction under 2 seconds, both orientations.

  7. One variable per test

    Never two at once.

  8. A proven control per network

    Not a number from a blog.

  9. Thresholds before launch

    Installs per variant, wait, KPI, margin.

  10. Custom events per step

    See exactly where players drop out.

  11. IPM to shortlist, value to scale

    Confirm with D7 ROAS or cost per purchase.

  12. Fatigue trigger and next variant

    Set in advance, built in advance.

How long should a playable ad be

Long enough to play the core loop once and feel a win. No network publishes a maximum length. Unity says teach the mechanic in 3 to 5 seconds, then let it set the length: 20 to 40 seconds for simple loops, over 2 minutes for puzzle, merge and mid-core (Unity best practices). Our genre guide maps mechanics to genres.

A test that worked

What a clean playable test looks like in practice

Judged on cost per purchase and Day 7 purchases, not on clicks.

Proof from our own work

DECA Games, DragonVale: 27% lower cost per purchase from playables

We ran three playable concepts for DragonVale on Google Ads, US Android users in English, over one month. Players hatch, breed and race dragons inside the ad. Against ad groups using only video and text assets, the playable ad groups cut cost per purchase by 27%, drove 3% more purchases on 12% less spend, and delivered 52% of Day 7 purchases from 44% of spend.

One title, one market, one month. It shows what a well built playable can do next to video. It does not promise the same result for another game.

Network lift figures are vendor claims, for example Moloco's unnamed advertiser with IPM up 49% and ROAS up 25% (Moloco). More in mobile game user acquisition.

FAQ

Playable ads testing: common questions

Short answers to the questions game UA and creative teams ask us most.

What is the most important metric for playable ads?

IPM is the most useful first read, but D7 ROAS decides. Unity's creative testing guide says to use early IPM to shortlist creatives, then let D7 and D28 ROAS plus retention settle the final decision. A playable can win on installs per thousand impressions and still lose on revenue if it attracts players the game does not keep.

What is a good IPM for a playable ad?

There is no reliable public IPM benchmark for playables by genre or network. Liftoff reports relative figures only: impression-to-install rates 8x higher than non-playables for top game spenders and 16x for other game advertisers in its 2025 index. Benchmark each playable against your own control creative on the same network and in the same country.

How long should a playable ad be?

A playable should be as long as its core loop and no longer. Unity says players must understand the mechanic within 3 to 5 seconds, because its close button can appear after 5 seconds. Simple, repetitive mechanics typically run 20 to 40 seconds, while puzzle, merge and mid-core loops can run beyond 2 minutes. No network publishes a maximum length.

How long should a playable test run?

Run a playable test until every variant has cleared the network's learning phase and a conversion threshold you set before launch. Unity says the learning phase usually takes a few days and warns against deciding within hours. A D7 ROAS read needs a 7-day-old cohort, and on iOS SKAN postback delays add up to six more days.

Is CTR a good metric for playable ads?

CTR is a weak metric for playables on its own. A persistent download button raises clicks by design, and taps during gameplay can be accidental. Read CTR next to CVR: high CTR with low CVR usually means accidental taps or a mismatch between the playable and the store page. Use IPM and post-install value to decide.

How does SKAN affect playable ad testing?

SKAN makes iOS playable reads slower and coarser. SKAN 4 sends up to three postbacks with delays of 24 to 48 hours for the first and up to 144 hours for the next two, and low-volume campaigns get coarse values only. Shortlist mechanics where data is richest, then confirm the winner on iOS at campaign level.

Sources. Checked on 19 November 2026: Unity: Ad creative testing guide; Unity: Playable best practices; Unity: Playable specifications; Unity: 5 myths about playable ads; Unity: Design playables in-house; Google Ads Help: HTML5 and playable ads; Meta for Developers: Playable ads; TikTok Ads Help: Playable ads; Mintegral Help Center; Mintegral data analysis guide; AppLovin: Creative guidelines; Google: SKAdNetwork guidelines; Adjust: How SKAdNetwork 4 works; PocketGamer.biz on Liftoff 2025 index; AppsFlyer, April 2025; Sensor Tower, October 2026; Moloco (vendor claim); SplitMetrics: SKAdNetwork guide; mobilegamer.biz: ASA rulings, 4 November 2026. Ad networks change their rules often, so confirm specs in each network’s own documentation before you upload.

Playable testing

Your next playable test should end with a clear decision

Admiral Media's creative team builds playables and runs the tests that judge them, from IPM to D7 ROAS. More than 500 million euros in ad spend managed, 200+ apps scaled.

Book a meeting

Talk to our team about playables

Tell us the game and the networks you buy on. A creative strategist replies within one working day.


Get in touch with us