Setting Up Your First Facebook Creative Test for a Game Prototype

A clean creative test needs four things in place before any money moves: a verified advertising account, working install tracking, one campaign with a separate ad set per creative, and equal budgets across them. Get those right and the result is readable. Miss any one and the numbers cannot support a decision.

Why this matters

The first test is the one most likely to be set up incorrectly, and its result is often used to decide whether a concept lives.

Setup problems do not announce themselves. The campaign runs, money is spent, numbers appear, and nothing indicates that the platform optimised toward one creative from the first hour.

Doing it in the right order takes an afternoon and makes the answer trustworthy.

Account and tracking prerequisites

Complete all of these before building a campaign.

1. A business account with billing verified. Verification can take days, so start early.
2. The app registered on the platform, with both store listings connected if you run on both.
3. An attribution or measurement partner connected, or the platform's own SDK integrated, so installs map back to campaigns.
4. The install event verified, tested from a release build, per Porting a Unity Prototype to a Publisher SDK.
5. Data handling settings configured for the platforms and regions you target, since misconfiguration here suppresses reporting.
6. A live store listing, even unlisted, because ads cannot send people to nothing.

The install event is the one to be strict about. Run a small manual test, install from the ad, and confirm the install appears in reporting before scaling anything.

If you cannot see one install arrive correctly, do not spend a budget finding out that none of them did.

Campaign structure for creative testing

The structure decides whether you learn anything.

One campaign. Objective set to app installs.

One ad set per creative. This is the important part. Creatives placed inside a single ad set compete, and the platform will allocate spend to an early leader, which produces a result about the platform's early guess rather than about your creatives.

Identical targeting across ad sets. Same markets, same broad audience, same placements. The creative should be the only variable.

Equal budgets, set at the ad set level. Campaign-level budget optimisation will move money between ad sets, which defeats the purpose.

Keep placements broad but consistent. If one creative is only eligible for some placements because of its aspect ratio, the comparison is compromised, which is why the aspect ratio check in The Pre-Test Checklist matters.

Budget split and duration

Two reasons for the duration. Platforms have a learning period during which delivery is unstable, and day-of-week effects are large enough to mislead over a short run.

Optimising for a deeper event sounds appealing and fails at this volume. There is not enough data for the platform to learn against it, so delivery becomes erratic.

Naming conventions that save you later

Boring, and it pays back within two tests.

Adopt a single pattern and never deviate. Something like: game, test number, creative name, market, date.

Three reasons it matters.

• Reporting exports are unreadable without structure, and you will be comparing across tests within a month.
• Creative names must be stable. The same hook tested twice should carry the same name, so you can track it across tests.
• Future you needs the date. Costs move seasonally, and a comparison without dates is misleading.

Keep a simple record outside the platform too: one row per test with the setup, the creatives, and the result. That record becomes your hook bank, per Reverse-Engineering a Winning Ad Hook in Four Beats.

Reading the results

Look at four things, in order.

1. Click-through rate, which tells you whether the creative stops people.
2. Click-to-install rate, which tells you whether the store page delivers on what the creative promised.
3. Cost per install by creative and by market, never blended.
4. Frequency, which tells you whether you have exhausted the audience rather than found a bad creative.

The relationship between the first two is the most informative part. A high click rate with a low install rate means the creative promised something the store page does not support, and that is a store page problem rather than a creative one.

Act only on clear separation. At small budgets a modest gap between creatives is noise.

Setup mistakes that invalidate the test

Six that recur.

• All creatives in one ad set, so the platform picks a winner in the first hours.
• Campaign budget optimisation left on, which does the same thing at a different level.
• Unequal aspect ratios, so creatives compete for different placements.
• Stopping early, during the learning period, and reading the result as final.
• Changing something mid-test, which resets learning and invalidates comparison.
• Broken install tracking, which produces a clean-looking report full of nothing.

The fifth is tempting and worth resisting. A test that is edited halfway through has produced two partial tests, neither of which is readable.

What we would do

Run a very small one-day smoke test first, with a single creative, purely to confirm installs arrive and attribute correctly. It costs little and it protects the real test.

Then run the real test with two creatives, equal budgets in separate ad sets, one market, four days, and no changes once it starts.

Write the decision rule before launching: what result continues, what result stops, and what result means inconclusive. Deciding afterwards is how marginal numbers become whatever you hoped for.

The short version

• Verify billing, tracking, and one real install before spending a budget.
• One ad set per creative, identical targeting, equal ad set budgets.
• Turn off campaign budget optimisation so spend cannot shift between creatives.
• Run at least three full days and change nothing mid-test.
• Optimise for install, not a deeper event, at this volume.
• Read click-through, click-to-install, cost by creative and market, then frequency.

Run a one-day smoke test before your first real test this week. If you want the campaign structure checked before it goes live, send us the setup.

Related reading: Creative Testing on a $500 Budget, The Pre-Test Checklist Before You Spend a Dollar on UA, and What Is CPI and How to Lower It.