Pre-launch creative testing: spend 3 to 5% of the launch budget on micro-tests to learn which angle, which CPI and which retention await you before signing the big cheque.
"We launch big, we adjust on the way." The launch budget of an app or a game is the biggest media cheque in the life of the company. And it is the one signed with the least information: nobody has yet put a real creative in front of a real user in the real country. So we adjust on the way. Except that on the way, CPI is at its highest, the ad platforms are learning on your first impressions, and every week of drifting costs the price of a quarter's worth of tests. My view: before spending the budget, spend 3 to 5% of that budget discovering what will make it fail. That is what a pre-launch micro-test is. It is not there to find the winning creative. It is there to stop you burning the budget on the losers.
Pre-launch creative testing: creative is half the result
Start with what you actually buy with a media budget. NCSolutions, in "Five Keys to Advertising Effectiveness" (August 2023), went through almost 450 consumer goods campaigns across digital and TV to break down what drives incremental sales. The answer: creative, 49%. Brand, 21%. Reach, 14%. Targeting, 11%. Recency, 5%. In other words, the lever teams spend the most meetings on, targeting, weighs four times less than the message. And it is the only one of the five you can test for a few thousand euros before launch day.
In app acquisition, the concentration is even more brutal. AppsFlyer, in its "State of Creative Optimization" report (April 29, 2025), analysed 1.1 million video creatives across 1,300 apps and $2.4 billion of spend between Q1 2024 and Q1 2025. In gaming, the top 2% of creatives absorb 53% of spend. Outside gaming, 43%. The biggest gaming advertisers produce 2,743 creatives per quarter to find that 2%. You will not produce 2,743. All the more reason to know, before launch, which of your ten or fifteen creatives stand a chance of joining that club. Because your launch budget will concentrate on them anyway. The only question is: will you know before or after spending it?
Even Supercell tests before launching, and still kills
Supercell is the studio that turned soft launch into a religion. Five live games, more than thirty projects stopped along the way according to its CEO Ilkka Paananen. Their method is precisely to put a game in front of real players in a few countries, read retention, and kill without mercy whatever does not hold. And yet. Squad Busters, launched globally in May 2024, made more than $100 million in seven months and 60 million downloads in a year. In October 2025, Supercell announced its closure, the first game in the studio's history shut down after a global launch, with this line in the statement: "what we learned was that we had got it wrong".
I take two things from it. First: if the most disciplined studio on the market still gets it wrong after months of soft launch, a scale-up launching without testing is not playing the same game, it is playing the lottery. Second: the test does not protect you from a bad product, it protects you from a bad budget. Squad Busters was expensive, but it was launched with known creatives, a known CPI and known retention. Its failure is a product failure over time, not a launch budget gone up in smoke on messages nobody had validated.
At the other end of the spectrum, Dropbox. In March 2008, before even having a stable product, Drew Houston posted a demo video on Digg. The waiting list went from 5,000 to 75,000 sign-ups overnight. Zero euros of media, one message, one audience, one measurement. That is the micro-test in its purest form: you check that the promise moves people before paying to broadcast it.
The protocol: 3% of the budget, three weeks, three questions
A pre-launch micro-test is not trying to optimise. It is trying to answer three questions your business plan filled in with assumptions.
- Which angle makes people act? Not which button colour: which message. Four to six genuinely different angles (competition, reward, belonging, mastery, novelty), each in two formats. You read IPM, installs per thousand impressions, not CTR. An angle that gets clicks without getting installs is a trap. Test ideas, not variants.
- What does an install really cost where you launch? The CPI in the business plan usually comes from an average benchmark. Liftoff and Singular, in their "Casual Gaming Apps" report (May 8, 2025), point out that iOS CPI is structurally higher than Android, with a gap above 4x in some genres: $21.03 for casino on iOS, $4.29 for RPG on Android. An average tells you nothing about your genre, your platform and your country. Three weeks of testing does.
- Do they come back? A real cohort, even a few hundred users, gives you a D1 and a D7. It is the only data that turns a CPI into an LTV, and therefore into a budget you are allowed to spend. Without it, you buy installs without knowing what they are worth.
My rule for the pre-launch micro-test
3 to 5% of the launch budget, no more. Two to three weeks. Four to six creative angles, two formats each. One real target country plus one mirror country with comparable CPMs. Three readings and a single decision: which angles go into production, which CPI we plan on, and whether retention allows the planned budget or forces it to be halved. If the test says no, you have just saved 95% of the budget.
The five mistakes that make the test useless
Testing in a cheap country. A one-dollar CPM teaches you what people who are not your customers want. The test happens in the launch country or in a country that resembles it in usage, CPM and regulation. I explained why in choosing your launch countries.
Testing without the platform's constraints. If your product is classed as restricted by Meta, Google or TikTok, test with the creatives that will pass moderation, not with the ones that will be rejected on launch day. The topic is covered in acquiring when ad platforms say no.
Stopping at the click. A pre-launch creative test is read on the install and on the day-one return, never on CTR. Platforms optimise for what you ask them: ask for players, not for the curious.
Testing only on social. If the launch plan includes out-of-home, geo-targeting around an event or TV, the message that wins on Meta and TikTok must be checked in a format with no sound and no scroll. For Essity, gamified social generated 14 million impressions and around 6 euros per lead while building a first-party base: it is this kind of mechanic, at small scale, that lets you test an angle and start recruiting a base of sign-ups before opening. It is what we are working on right now with a competitive gaming platform: which angles get a player to sign up before the game is even open, and which ones only get a click.
Running the test the week of the launch. A test that leaves no time to reproduce the winning creatives is worthless. Three weeks of testing, two weeks of production, then launch. Five weeks of delay that avoid three months of wandering.
What I tell founders who want to "launch big"
Launch big, yes. But launch big on validated messages, at a measured CPI, with known retention. The micro-test is not caution, it is the only way to know how much your launch budget is allowed to weigh. It costs 3% and decides the remaining 97%. Teams that skip it do not gain three weeks: they lose them three times over during the launch, when every impression costs the most and everyone is watching.
If you are preparing an app or game launch in the next six months, go through our free audit: we look together at which angles deserve a test, in which country, and at what budget. And to visualise each channel's share in your plan, the media mix simulator is there for that.
