012024
Regulatory environment assessment
Policy & advocacy research • Qualitative analysis
Donor reporting • Pipeline automation
Four donors, four reporting templates, and the same underlying figures being re-derived by hand for each one.
- Client
- Confidential — independent engagement
- Role
- Analytics engineer
- Duration
- 2 months
- Year
- 2025
- Python
- dbt
- GitHub Actions
The challenge
Because each template was built separately, the same indicator could end up different across two reports submitted in the same week — usually a filter someone forgot to apply.
The approach
One modelled results layer, with each donor template as a thin projection over it. The figures are derived once; the templates only reshape them.
Generation runs on a schedule and fails loudly when an indicator is missing, rather than emitting a blank cell.
The outcome
Reports agree with each other by construction, and the reporting week stopped being the busiest week of the quarter.
Results
- Sources of truth
- 4 → 1Sources of truth
- Template generation
- AutomatedTemplate generation