In 2025 we published a game that we pulled from sale after six weeks. This is the write-up. I am publishing the numbers because when we were making the decisions that caused it, nobody else in this market had published theirs, and we made every mistake that a public post-mortem would have prevented.
The first error was pricing. We set a single global price with a regional discount applied on top, which meant the Indian price was derived from a US number rather than from what the market supports. The result was a store page that converted at 0.9 percent in our largest addressable market against a 3.4 percent benchmark, and we spent three weeks blaming the trailer.
The second error was the UA plan. We bought installs against a lookalike audience built from a completely different genre because that was the audience data we had access to. Cost per install looked acceptable. Day-seven retention was 4 percent. We had bought the cheapest possible wrong players and the dashboard told us it was going well for eleven days.
The third error is the one I think about most. We had a beta cohort of 400 people who told us, in writing, that the progression was too slow. We interpreted that as a tuning note and shipped a tuning change. It was a structural note. The progression was not slow, it was illegible, and no amount of multiplying numbers was going to fix a system nobody could describe back to us.
What we changed: pricing is set per market from purchasing power before anything else is decided, UA audiences are built from our own genre data or not at all, and any beta feedback that appears more than twenty times gets a structural review rather than a tuning pass. Three titles since, none pulled.
