We test many creatives cheaply, then scale the few that win
- 01Test many creatives cheaply
- 02Scale the 4 of 20 that win
- UGCstreamer reaction16% D7 ROAS$5.1k spend
- Reelboss fight14% D7 ROAS$4.2k spend
- Reelco-op moment12% D7 ROAS$3.6k spend
- UGCfirst ten minutes11% D7 ROAS$2.9k spend
- Statickey art
- Carouselcharacter roster
- UGCplayer review
- Staticstore rating
- Reelcinematic trailer
- UGCtier-list rant
- Staticupdate notes
- Carouselmap tour
- UGCbuild guide
- Staticlaunch reward
- Reeldev diary
- Staticpress quote
- Reelpatch trailer
- UGCPvP clip
- Carouselskin showcase
- Staticsale banner
Illustrative test for a free-to-play mobile game, optimised to in-app purchases. Most creatives are a cheap probe. The 4 that beat the day-7 ROAS bar get the budget concentrated on them; the other 16 stay small and get cut.
What we run inside your Meta account.
Account and campaign structure
Run a consolidated build: two to four campaigns, eight to fifteen creatives per ad set, Advantage+ where volume allows, and manual ad sets for launches, new countries, or restricted categories.
Audience strategy
Run broad Advantage+ audience, keep existing users and players out of prospecting, and test lookalikes of paying users against that baseline. Teens can be reached only by age and location.
Creative — the new targeting
Meta's Andromeda system reads creative to find responsive users, so ten to fifteen conceptually distinct concepts beat near-duplicates suppressed under one Entity ID. Every concept faces AI-simulated focus groups before it spends; live, we diagnose hook, hold, and click, then cost per paying user.
Pixel and Conversions API
Run Pixel and Conversions API together, deduplicated by matching event ID within 48 hours. Send backend events like trial activation, paid plan, or a funded account with hashed first-party data; mobile games send in-app events through Meta's SDK or an MMP.
Offer, funnel, and landing page
Match the ad's promise to the landing headline, signup flow, and app-store page. Advantage+ can't repair a signup flow people abandon.
Retargeting, testing, and scaling
Retarget by stage (started a trial but never activated; installed but never bought), test concepts separately from variations, and graduate winners. Most new scale comes from new concepts.
Every creative plotted by spend and ROAS — the winners rise above the line
Illustrative: the same 20 mobile-game creatives, each placed by spend and day-7 ROAS and sized by installs. The dashed line is a 10% day-7 ROAS bar, set from the game's payback curve. The 4 above it took the budget; the rest stayed at probe spend and were cut.
Who does what on your Meta account.
What we handle
- Setup and reconciliationPixel and Conversions API with verified deduplication, consent gating, and reconciliation to activated trials, in-app purchases, or Stripe revenue.
- Strategy and buildAudit, 90-day plan, Advantage+ and manual structure, app campaign setup, and exclusions.
- Creative directionBriefs for product demos, founder explainers, gameplay capture, creator reactions, onboarding walkthroughs, and customer quotes.
- Testing and optimizationSingle-variable creative tests, winner graduation, and weekly fatigue checks.
- Reporting and incrementalityPlatform ROAS beside blended MER from your backend, plus holdouts for incremental lift.
What we need from you
- AccessAdvertiser partner access you can revoke at any time. You keep the ad account, Pixel, dataset, and audiences.
- BudgetMedia spend set in your written scope, billed by Meta, never marked up.
- Creative accessA test account or game build we can screen-record, plus founders, developers, or customers willing to appear on camera.
- Licenses and approvalsAny approval your category needs, like Meta's written permission for crypto or gambling authorization for real-money games, plus timely creative and compliance review.
- Weeks 1–2Audit and foundationVerify deduplication, align attribution with your backend, file any written-permission request, ship the first creative, and deliver a 90-day plan.
- Learning phaseExit the algorithm's learningEach ad set needs about fifty optimization events in seven days. At $40 an event that is about $2,000 a week, or $8,700 a month per ad set. Targeting changes, big budget swings, or pauses of a week or more restart learning.
- OngoingPerpetual creative refreshProspecting creative wears out. We stage new concepts before frequency climbs or CPM rises.
What you get.
| Deliverable | What it is |
|---|---|
| Account, Pixel, and Conversions API setup | Verified deduplication, event-match baselines, consent gating, and backend reconciliation before spend. |
| Creative brief pack | Ten to fifteen distinct concepts per campaign, from product demo to gameplay clip, each with its hook, hold, and CTA specified. |
| Ad-set and campaign build | Advantage+ and manual split, prospecting exclusions, and a retargeting ladder. |
| Creative-testing plan | Equal-budget concept and variation tests with winner rules and a scaling path. |
| Weekly report | Platform ROAS beside blended MER from your backend, plus CPA, CTR, CPM, frequency, hook rate, and fatigue flags. |
Four rules for Meta spend.
Feed the algorithm signal, then let it target.
Give Meta clean, deduplicated events and first-party data, then let the auction find responsive people.
Distinct angles beat polished duplicates.
Andromeda reads creative and collapses near-identical ads under one Entity ID. The edge is distinct angles, like a boss fight, a creator reaction, and a co-op moment, over one glossy trailer cut ten ways.
Every winner decays — refresh is perpetual.
Winning ads depreciate across frequency, hook rate, and CPM. Stage the next concept before fatigue stalls the account.
Report the honest number, not the flattering one.
Meta's default attribution counts conversions up to a day after a view, including some that would have happened anyway. We show platform ROAS beside blended MER reconciled to your backend and test lift with holdouts. Budget follows the honest number.
Start with your paying players, then expand outward into lookalikes
- Seed audienceplayers who made a purchase
- 1% lookalike0.4M reach
- 3% lookalike1.2M reach
- 5% lookalike2.0M reach
- 10% lookalike4.0M reach
Illustrative, for the same mobile game in one country of about 40 million people. A seed of players who made a purchase sits at the center, and Meta builds lookalikes outward at 1%, 3%, 5%, and 10%. Tight percentages stay closest to the seed. Each ring runs against a broad Advantage+ audience, and budget follows whichever finds payers cheaper.
How the trend engine feeds Meta ads.
The trend engine is our weekly read of search, social, and launch signals. We brief the hooks and formats it shows winning in-feed, then test every variant on simulated panels, before your competitors' media buyers notice them.
How our models are trainedQuestions buyers ask
- What budget do I need to start, and to scale?
- Plan for about $8,700 a month in media per ad set at a $40 cost per event, plus our fee from $1,500. That buys the fifty events a week an ad set needs to exit learning. We usually start with one or two ad sets; if your event costs more, we optimize to an earlier one, like an install or trial start.
- How many creatives do you need each month, and why?
- Enough to keep ten to fifteen distinct concepts live with replacements staged. A new angle, like a boss-fight clip, counts for more than a new cut of an old ad.
- Can you run Meta ads for a mobile game?
- Yes. Advantage+ app campaigns optimize to installs, an in-app event, or purchase value once events arrive through Meta's SDK or your MMP. iOS results come back through SKAdNetwork or Aggregated Event Measurement, later and in aggregate. Games where players stake and win something of monetary value need Meta's gambling authorization.
- Can you run Meta ads for a crypto product?
- Yes, within Meta's policy. Exchanges, crypto lending, and wallets that buy, sell, swap, or stake need Meta's written permission, backed by a license Meta recognizes, such as FinCEN MSB registration or a New York BitLicense in the US, FCA, BaFin, or MAS. Education, news, and storage-only wallets don't.
- Do you test creative before it spends?
- Yes. We generate variants of each concept and score them on AI-simulated focus groups, run in aggregate at scale, for hook, CTA, the gestures that point at it, and implied meaning. Reinforcement learning tunes the next round toward what the panels reward, and only the strongest variants launch.
- Who owns the ad account, the Pixel, and the creative?
- You own all three. We use Advertiser partner access you can revoke, the Pixel stays in your name, and the contract assigns creative to you. Ad-account ownership can't move between businesses, so an agency-owned account is a hostage risk.
- What did iOS 14 and privacy changes do — did they break tracking?
- They degraded it. Conversions API data restores aggregate signal, but it can't override an ATT opt-out; opted-out iOS app attribution still runs through SKAdNetwork. Claims that CAPI restores individual iOS tracking are false.
- Why is Meta's ROAS different from my real numbers?
- Meta credits conversions up to seven days after a click and one day after a view, and it can't see refunds or trials your backend later cancels. Use platform ROAS directionally; budget against blended MER, revenue divided by total spend, from your backend, with holdouts measuring lift.
- How long until it is profitable?
- Setup takes weeks one and two; first winners may show in month one, with scaling and an incrementality read in months two and three, inside the three-month initial commitment.
- Can you guarantee a specific ROAS or result?
- No. Results depend on offer and economics. We commit to clean signal, distinct creative, and reporting that separates Meta's claims from what your backend confirms.
Scale on creative. Count what settles.
We'll audit account and Pixel signal, find spend leakage, check what Meta's policies for your category let you run, and put the blended number next to the platform number.