drive_wave’s first-week postmortem for Coin Harvest puts sales, reviews, and refunds beside the bugs players actually encountered. The desktop idle game’s store page disclosed AI-generated promotional art and some in-game assets. Its developer argues that disclosure reduced players’ tolerance for flaws, while acknowledging substantial launch problems.
- Sales and reviews diverged. The author reports roughly 6,900 copies sold, but a review score that fell from 83% to 55% positive. The latter rested on 49 reviews.
- Refunds varied by market. An 11% overall rate concealed 5–6% in the US and Japan and much higher rates elsewhere. The developer says regional refunds exposed localization problems before written reviews did.
- The complaints had concrete targets. Chainable time warps broke progression, coins became unreachable, upgrades felt weak, and realms felt repetitive. Cloud saves were also missing at launch.
- Background runtime clouds engagement. Reported median playtime was 2.5 hours, with a six-hour average. A widget left open while someone works can accumulate hours without demonstrating satisfaction.
- Arguing made support worse. The author’s revised approach is to reply once, explain the fix, and move on. Their other priorities are polishing the core loop and tracking regional refunds from day one.
Break launch feedback into defects, audience expectations, and buyer groups. That makes a poor score actionable: a localization problem needs different work from a progression exploit or a store page attracting the wrong audience.
These are self-reported results from one game; they cannot isolate AI disclosure’s effect from bugs, audience mismatch, or changes in who bought it.