Approach
How we did it, rather than how we do it.
A methodology described in the abstract proves nothing. Here are the four stages as they actually ran on the Eco Afrique system.
- 01
Scoping
30 minutesOne call. We identify the real problem, not the solution you have already imagined. If the project does not stand up, we say so here.
For Eco Afrique, the stated problem was "digitise job weighting". The real problem was: nobody knows whether a dossier is complete until they have opened all of it.
- 02
Scope and quote
3 to 5 daysA written scope, a firm price, a schedule. What is in and above all what is out — the part that prevents conflict in month three.
The weighting module started with a constraint written into the scope from day one: no personal data. That determined the entire architecture afterwards.
- 03
Build
4 to 10 weeksVisible deliveries every week on an environment you can open. You watch the system grow rather than receiving a reveal at the end.
Some weeks produce a screen. Others produce a database migration and nothing visible. We say which of the two is happening.
- 04
Launch, then after
ongoingCutover, documentation, handover of access. Then upkeep: alerts, backups, fixes, evolutions.
Eco Afrique has been in production since June 2025. The most recent documents processed are from yesterday. That is the part most providers do not show, because it does not exist for them.
What we build on
We do not change stack every project. A small, well-known toolset produces fewer surprises than a fashionable one, and is easier for someone else to take over.
- Interface
- React, TypeScript, Vite
- Backend and database
- Supabase, PostgreSQL, row-level security
- Server functions
- Deno edge functions
- Automation
- self-hosted n8n
- Applied AI
- OCR, structured extraction, assisted evaluation
- Public sites
- Astro, static generation
Communication cadence
A weekly check-in
Short. What shipped, what is blocked, what is next. Written, so it stays consultable.
An open staging environment
You can click through it whenever you like, without telling us.
Bad news first
A delay announced early is an adjustment. Announced late, it is a breach of trust.
What we need from you
A project rarely fails on its own. Here is your part, stated in advance.
One reachable decision-maker
A person who decides. Not a committee that approves in three weeks — the single biggest cause of schedule slippage.
Your data and content on time
A system can be ready and sit unusable for two months for lack of content.
Batched, fast feedback
On the weekly deliveries. Feedback scattered across three weeks costs more than the development itself.
Shall we look at your project?
Thirty minutes, no commitment. By the end of the call you will know whether it stands up and roughly what it costs.