Kobo Count v3 is in review. What I learned shipping to low-end Android

Three apps in the stores, and v3 took four months because of phones I do not own: 1GB RAM, Android 9, intermittent everything. Lessons:

Offline first is not a feature, it is the app. Every screen works with stale data and syncs silently. The sync queue survives process death (learned twice). APK size under 15MB or downloads fail on market wifi (dropped from 38MB, mostly images and one foolish font). Battery: the background sync that polled every 15 minutes now runs on charging + wifi only, complaints stopped.

Review tip that actually worked: a 40-second video of a trader using it at the market, in the listing. Conversion from listing to install doubled.

THE VIDEO TIP!!! We did the same for Habitloop, real user on camera, installs jumped. Store listings are a conversion page and everyone treats them like a form. What is your listing conversion now?

Listing-to-install 31%, was 16%. Same screenshots, added the video and rewrote the first two lines. First two lines are the whole listing, the rest is decoration.

Low-end Android solidarity. Salah Times Pro supports Android 7 and the crash logs from 2016-era devices are a genre of literature. Question: how do you test without the devices?

One used Tecno bought in the market for 45 dollars, plus Firebase Test Lab on the cheapest profiles before every release. The 45-dollar phone finds more bugs than the lab.

15MB is brutal and correct. What got cut to get there?

The foolish font, three onboarding illustrations replaced with one, all PNGs to WebP, and one SDK that was 9MB for a feature 2% used. The SDK was the crime.