I'm a systems engineer in building automation. Fazor is a side project: eleven electrical calculators for electricians in one offline Android app. It went live on Google Play on 18 November. Here is what the launch looked like, with the parts that surprised me.
The product
Cable size, voltage drop, breaker and RCD selection, motor starting current, power factor correction and a few more. IEC or NEC, the formula and the standard reference shown under every result.
Kotlin and Jetpack Compose, single activity, Hilt, Room and DataStore. No backend, no account, no analytics. The calculation code is plain Kotlin kept apart from the Android framework, so it runs under ordinary JVM unit tests: 397 of them, with 98.5% line coverage on the calculation logic. Rules that touch life safety, such as RCD selection, are written as explicit tables with one test per rule.
Five calculators are free with a banner ad; a one-time US$2.99 unlocks the rest and removes ads.
The 12 testers x 14 days rule
New personal developer accounts must run a closed test with at least 12 testers opted in for 14 continuous days before they can apply for production access.
- Finding 12 testers took one day: my own network plus free tester communities.
- The 14 days were quiet. I got no written feedback from testers at all, even after sending a five-question survey on day 13.
- I expected that to sink the application, since "insufficient tester engagement" is one of the two stated rejection reasons. I answered the form honestly: no tester reported a wrong result or a crash, and the app has no analytics, so I cannot see what they used.
- Production access was granted the same day I applied. The production release was approved and live the same day too.
So the rule cost me 14 days of waiting, not 14 days of work. If you are about to start: put the feedback questions in the invitation on day one, not day 13.
One practical trap: the Play Console form limits each free-text answer to 300 characters, which the help page does not mention. Draft short.
What I did during the wait
No new features. The decision was to ship exactly what testers had, and to release an update only if someone reported a bug. Nobody did. Instead I measured test coverage and made the promo video.
A promo video made in code
I had no budget for video and the AI video tools I tried were locked behind paid plans. So the video is a web page.
- A 1080x1920 stage in plain HTML and CSS. Every motion is a CSS animation with
animation-play-state: paused. - A
seek(ms)function setscurrentTimeon every entry ofdocument.getAnimations(), so any frame can be drawn on demand. - Puppeteer drives an installed Chrome, takes a screenshot per frame at 30 fps and pipes the PNGs into ffmpeg.
- A small Python script synthesises the motion sound effects, places the text-to-speech narration, stretches each scene to fit its line and writes the timing file the page is built from. Scene length follows the voice, not the other way round.
- Turkish and English versions come from one content file.
The result is deterministic: the same inputs give the same video, and changing a caption is a one-line edit followed by a re-render.
It took a few iterations. The first cut was boring, the synthesized music was bad and was replaced by a licensed free track, and a decorative rotating background got cut after one viewing.
Numbers so far
Launched this week, so there is nothing worth reporting yet. I will post an update once there is a month of data.
What I would do differently
- Ask testers for feedback from the first day.
- Check that my own product page did not still say "coming soon" on launch day. It did, for a day.
- Create the brand accounts before launch instead of on launch day.
Happy to answer questions about the Play Console process, since most of what I found online about the tester rule was guesswork.
Product page: https://kalkansoftware.com/fazor/
