Choosing soft-launch geos: the two-rung ladder

A soft launch exists to answer two questions before global spend makes them expensive: does the game keep people, and does it make money per person it keeps. Most soft-launch failures are geo-selection failures in disguise, because the two questions want opposite markets. Retention needs volume, which favors cheap installs; monetization needs launch-market-like wallets, which cost real money. The ladder answers both by refusing to pick one market.

Rung one: the volume geo

Purpose: technical stability, funnel reads, early retention shape, and creative testing at survivable prices. Candidates share three traits: low CPI, large Android populations, and store review speed you can live with. The Philippines, Vietnam, Indonesia, and Brazil are the perennial picks, with CPIs that run far below tier-1 rates. What rung one can answer: does onboarding leak, do sessions crash, does the core loop hold a D1 in a healthy band. What it cannot answer: whether anyone will pay launch-market prices, because you deliberately bought users who mostly won't.

Rung two: the signal geo

Purpose: monetization truth. You want a market whose players behave like your launch market's, at a tolerable sample cost: the Netherlands, the Nordics, Canada, New Zealand, and Australia are the usual suspects, wealthy, English-friendly or near enough, and small enough that honest samples don't bankrupt the study. Expect to pay tier-1-adjacent CPIs here and buy fewer users; that's the price of reading ARPU, purchase timing, and ad-revenue-per-user numbers you can extrapolate without lying to yourself.

The gates, written before the first euro

The ladder only works with pre-committed gates. Ours follow one shape: a D1 retention floor set against the genre's published band (all-vertical averages run about 26% at D1 and 13% at D7, with genre medians like puzzle's 31.85% D1 as the honest bar), a CPI ceiling at soft-launch scale, and an early revenue slope the payback model accepts. Gates are written down, dated, and signed before launch, because gates negotiated after data arrives always lose. Thepuzzle post-mortemshows a gate refusing a global launch by one point, correctly.

The classic misreads

• Reading Manila retention as Munich revenue. Volume-geo users retain differently and pay differently. Never extrapolate rung-one monetization; that's rung two's job.
• Testing creative in one geo, launching with it globally. Hooks transfer across borders better than endings and CTAs do; re-validate winners in the signal geo before trusting them at launch.
• Letting the volume geo flatter blended numbers. Report the rungs separately, always. One blended CPI across both rungs is a number with no meaning and great confidence.
• Store-listing neglect. Soft-launch conversion rates carry your store page's fingerprints. If the icon and screenshots are placeholders, the CPI you measure is a placeholder too.

When the ladder says no

The most valuable soft launch is the one that cancels a global launch. If D7 won't clear its floor after honest iteration, the auction won't fix it: install costs only rose this year, 15 to 20% on the big social channels, and buying growth for a leaking game gets less forgivable annually. The kind thing, to the game and the budget, is to send the build back to the designers with the cohort curves attached. The ladder's job is to make that verdict cheap, early, and undeniable.