Field notes: a week at a decentralised trial site

We spent last week on-site with a study team in Gaborone running a UTI-diagnostic trial. Not on a demo call. Not walking a room through a mockup. Actually watching the clinical officer and lab technician use the tablet, all day, in the flow of real appointments.

Five things stood out.

1. Battery, not bandwidth

We spend a lot of time optimising for low-connectivity clinics. Turns out the harder constraint on this trip was the tablet running out of charge by 3pm. The lab tech ended up leaving the app open all morning; our screen never dimmed. That's on us.

Fix in build: aggressive auto-lock + a "start-of-shift" recap on resume so the tech doesn't lose their place.

2. Everyone wants one more scan

The pack-mismatch flow (wrong patient / wrong site / pack already used) worked exactly as designed — twice in five days it caught what would have been a chain-of-custody break. The team asked us to add one more scan step at the courier hand-off. Their instinct, not ours.

3. Reading test strips is still hard

The Cube Reader flow is meant to remove the "guess the shade of red" problem, but we watched two techs disagree about a borderline strip. The scorecard reference card helps. A photo comparison side-by-side with the strip would help more.

4. Language switching is not a nice-to-have

French translations are still marked "pending from client" in the build. They're not pending anymore — a French-first tech had to flag mentally on every screen. Priority #1 for the pilot cut.

5. The paper backup didn't get used once

That's the win. Old workflow required carbon-paper triplicate for every sample. New workflow: scan, done. The pad stayed in the drawer.

The best thing about being on-site is that the polite feedback you get in a demo room evaporates. Everything is honest. Everything is urgent. If you build clinical software and can go — go.

More field notes to come as we roll out to Cameroon next month.