Five years of Bird Sound Identifier
2,100 sales later — what a niche ML app taught me about pricing, reviews, and listening to users (and birds).
Bird Sound Identifier turned five this spring: 2,100 sales, 290 reviews, a 4.8★ average, and a support inbox that is mostly people sending me recordings of very confused starlings.
The lessons were not the ones I expected. Pricing mattered less than I feared — the people who want on-device birdsong recognition will pay for it, and the ones who don't were never customers. Reviews taught me more than analytics ever did: almost every feature worth shipping started as a one-star complaint written at 6 a.m. from a forest somewhere.
And the model is the least interesting part. BirdNET does the heavy lifting; the work is everything around it — microphone handling in wind, offline species ranges, making a result you can trust when the app says it heard a wryneck. Five years in, that's still the job: not identifying birds, but earning the benefit of the doubt.