Your competitors' angriest customers already wrote your spec
Before I wrote a line of the new Fidelia, I read reviews. Not the five-star ones — those tell you what people tolerate. The one-star reviews, years of them, for our app and for every music player in the category.
I expected feature demands. That is not what one-star reviews are made of.
The anatomy of one star
People do not leave one star because something is missing. They leave it because something was taken. The update that scrambled a library that took a decade to tag. The app that forgot the window layout, again. The plugin that crashed the player and took twenty minutes of listening with it. The subscription that appeared where a purchase used to be, holding a customer’s own files hostage for a monthly fee.
Read enough of them and the category’s real contract emerges, and it has almost nothing to do with feature lists. It reads like this: never lose my library. Never crash. Never make me rent what I already bought. Remember where I put things. When something goes wrong, act like it matters.
Boring promises. Unmarketable promises — no App Store screenshot says “doesn’t lose your stuff.” And they are the entire product, because a music library is one of the few pieces of software state people carry for decades. Betray it once and the review writes itself.
What that did to the roadmap
The rebuild’s priorities came straight off that reading, in order.
The library is sacred: the importer and the database got deterministic test harnesses before the UI got its first coat of paint, and we stress every library feature against a synthetic hundred-thousand-track library before it merges, because the customer with the biggest library is the one who loves you most and has the most to lose.
Crashes are contained, not managed: third-party plugins run out of process, so somebody else’s bug kills a helper process instead of the music. This month that architecture turned five crash logs from one machine into a same-day, one-line fix.
The business model is a purchase: pay once, own it. The category taught everyone what the alternative earns.
And the small betrayals get treated like large ones. A customer wrote that his windows drifted a little on every launch. We measured his screen recording pixel by pixel before answering — because the review that starts “the developer didn’t believe me” is the one that costs you the next hundred customers.
The general case
None of this is specific to audio. If you are building into a category with incumbents, their angriest customers have already documented, in public, for free, exactly where the incumbents break their promises. Not where the features are thin — where the trust is. That list is your spec, and it is sorted by pain.
The discipline is reading it before you fall in love with your feature list, because the reviews will not care about the feature list. Nobody ever left one star because an app had too few features. They leave it because the app broke a promise it never knew it was making.
Fidelia is on the Mac App Store — free trial, one-time purchase, and a library that is treated like it belongs to you, because it does.
If you are entering a category and want the incumbents’ broken promises turned into your roadmap: that is the day job. audiofile.engineering/engineering.