raymond-UI/fsqca-african-fintech

# fsQCA — African Fintech Funding Configurations A structured, auditable pipeline for running a fuzzy-set Qualitative Comparative Analysis (fsQCA) study end to end: from raw evidence gathering on individual startups, through calibration and analysis, to a written paper draft. **Research question.** What configurations of founder characteristics and network ties are associated with higher seed-stage funding among African fintech startups (2015–2024)? This repository is designed to be operated by an automated agent, one task at a time, with a human reviewer approving at three gates. Every fact the study rests on is stored as a structured record with a source pointer, so the final paper is fully traceable back to evidence. --- ## How to operate this repo (read this first, agent) 1. **Never invent data.** If you cannot find a fact after a genuine search, record it as `status: not_found, value: null`. A missing value is a valid, useful result. A guessed value silently corrupts the entire analysis. This is the single most important rule in the repository. 2. **One task at a time.** Work the task list in `PROGRESS.md` top to bottom. Each task has defined input files and exactly one output file. Do not start a task whose inputs are not yet `done`. 3. **Stay inside your stage.** Evidence-gathering tasks (C1–C4) search the web and write facts. The coding task (C5) may *only* read what C1–C4 already wrote — it never searches fresh. This guarantees every score traces to recorded evidence. 4. **Update your own cell.** When you finish a task, set its cell in `PROGRESS.md` to ✅ and commit. Use one commit per completed task, message format: `[case-id] Cxx: short description` (e.g. `flutterwave C2: funding rounds`). 5. **Stop at gates.** Do not cross a 🔴 GATE row in `PROGRESS.md`. A human sets those. If you reach one, stop and summarise what is ready for review. --- ## Repository map ``` . ├── README.md You are here. Agent operating manual. ├── STUDY_DEFINI …