Field notes

Field notes: what early testers in Uganda changed in our products

Qualitative observations from early testing and community sessions in Kampala — the assumptions that broke, and what we changed because of them.

Why we publish this

We say community feedback shapes our tools, and that claim deserves receipts. These notes are qualitative — observations from early testing sessions and structured feedback from pilot teams — and they are published in that spirit: honest about what broke, specific about what changed.

What we observed

First: mobile-first was not a preference, it was the default reality. Testers arrived on Android phones first and laptops second, often on constrained data. Features that assumed a wide screen or heavy assets were quietly abandoned, regardless of how useful the scoring underneath was.

Second: vocabulary decided trust. Words like 'algorithm' and 'model' closed conversations; words like 'checklist', 'score with reasons' and 'draft for your review' opened them. The capability barely changed between those framings — the willingness to test it changed completely.

Third: the CSV upload was the feature. Teams that 'did not have data' turned out to have years of it in spreadsheets and exercise books. Meeting teams at their actual data reality produced more adoption energy than any connector roadmap we described.

What we changed

We rebuilt primary flows around a 320-pixel baseline before adding desktop polish. We rewrote result screens so every score arrives with its stated reasons in plain language. And we kept the CSV/Excel-first path as a permanent design decision rather than a temporary bridge, because for our users it is not a compromise — it is the product.

Actions you can take

  • If you are building for African markets: test on the phone your users carry, on the connection they pay for.
  • Rewrite one screen of your product in the plainest language you can and watch what happens to completion.
  • Ask users where their data really lives before designing your import experience.

Limits — what stays uncertain

  • These are qualitative observations from a small early group, not survey statistics — we publish them as direction, not proof.
  • What we observed in Kampala will not map perfectly onto every market; treat local testing as non-negotiable.
Keep reading

More from Resources

Insights6 min read

AI readiness is a people problem first

Most organizations asking about AI are not blocked by technology. They are blocked by unclear priorities, unmapped workflows and unspoken fear. Readiness work starts there.