Client
A French start-up, marketplace app on web and mobile
Team
A development agency in Vietnam, fully remote
My tools
Notion
Where it stood
The app was due in January 2025. When I arrived in July, two core features still didn’t work, and the founder wouldn’t onboard users until they did. More than 90 tickets were waiting across two Trello boards, under 20 different statuses. The founder couldn’t see what was being built, and the agency couldn’t see why. Neither side lacked skill. What was missing was someone translating between them.
My role
Digital project manager between the founder and the agency. I spoke with her in French, turned her requests into tickets the team could build without guessing, recorded short videos when a feature needed showing, and wrote release notes after each delivery plus a monthly report for her. In eight months, 18 releases and about 150 tickets went out.
How it went
- Sep 2025 — The agency’s project manager left, then its product owner and tester, who weren’t replaced. A fixed price no longer protected anyone, so the remaining work moved to a monthly budget.
- Nov 2025 — Search went from about 30 seconds to 3. The fixed-price scope stood at 60%, with 33 tickets delivered since September.
- Dec 2025 — End-to-end demo of the fixed-price scope. The app ran, with too many bugs to launch.
- Jan–Feb 2026 — Weekly delivery every Tuesday. But bugs fixed one week came back the next, and the question on the table became: fix the existing app, or rebuild it?
Why it stopped
In March 2026, a prototype rebuilt with AI in about a week reproduced most of the existing platform. The founder’s business model had evolved too, and she chose to restart with as few people as possible. She ended the contract that covered my role and asked me to stay on in a different one, documentation and marketing. That isn’t the work I do, so I handed over on 13 March and stopped there.
What I’d do differently
- Push for a monthly budget from the first month, not the third. On a product whose scope still moves, a fixed price makes everyone defend the scope instead of the product.
- Ask for automated regression tests before the first release, not after the tenth. A bug that comes back is a testing problem before it is a people problem.
- Spend the founder’s time on the why and the what. In 2026 a prototype costs a week. What stays expensive is knowing exactly what to build, and for whom.
Leave a Reply