Thanks, and agreed. The in-app goal was the lesson, and I'd rather have learned it at $220 than at a trial-run budget. The part I found interesting is that Google's count wasn't wrong: there really were 21 installs. You only saw the problem by looking at which version of the app they installed. The YC advice is roughly where I've ended up too. And fair point on the refund, I'm not holding my breath, but I'll report back either way.