How Do You Know It's Time to Spend Real Money on User Acquisition?
A soft launch feels like a smaller launch. It isn't — it's a decision gate wearing a test's clothes. Done right, it tells you whether to pour budget or kill the project; done wrong, it's just an expensive way to collect screenshots. Here is what the gate is actually checking.
Q: What is a soft launch actually supposed to validate?
Not "does the game run" — that's QA, and you should know before you spend a cent on traffic. A soft launch validates unit economics and fit: the shape of your retention curve (D1, D7, D30 — not just install), whether LTV clears CPI inside a payback window you can survive, and which regions actually pay. It's the difference between "people downloaded it" and "this is a business." If your soft launch deck is full of install counts and empty of cohort LTV, you launched a vanity metric.
Q: Which signals must be green before you pour budget?
Three, and they're gates not vibes. D1 retention at or above your category benchmark. A payback period short enough that you don't go underwater — many teams use sub-90-day as the line, adjusted for how much runway they have. And a cohort LTV projection you can defend, meaning it's built from enough days of real data, not a three-day extrapolation wearing a trend line. The counterintuitive trap: scaling early buys noise (your LTV error bars are still huge), scaling late cedes the window to a clone. The discipline is a hard threshold, not a feeling.
Q: Why do so many teams scale at the wrong moment?
Vanity metrics — installs going up looks like success, so they spend. Funding rhythm — a board meeting next month pushes a "growth story" that isn't real yet. And region blindness — they treat the soft-launch country as the whole world, ignoring that payment rails, price sensitivity, and cultural fit differ per market. None of those are data. The teams that scale well treat the threshold as contractual: below it, you don't spend no matter how good the pitch deck looks.
Q: What breaks first when you actually scale?
Almost never the user side — it's the system side. Server capacity that was fine at 50kDAU falls over at 2M. Support SLA collapses. Payment compliance in the new region needs KYC and tax registration you didn't file. Localization has gaps exactly where you just bought users, so they land on untranslated screens. These non-user bottlenecks are invisible during soft launch and erupt the moment you pour gasoline on it. The pre-scale checklist is operational, not creative: load test, support plan, payment licenses, localization coverage.
| Signal | Gate | If red, don't scale |
|---|---|---|
| D1 retention | category benchmark | users don't return |
| Payback window | survivable (e.g. <90d) | you go underwater |
| Cohort LTV | defensible from real days | guessing, not modeling |
| System readiness | load/support/compliance | you break on contact |
