Portfolio case study: reproducible data architecture and multi-source quality controls for anticipatory action in Niger.
English · Español
# Case study · Data for anticipatory action in Niger
> From fragmented humanitarian sources to a traceable, reproducible data architecture designed for multi-source analysis.
**Author:** Marta González Vázquez
**Context:** internship at Action Against Hunger Spain · Digital Transformation
**Role:** data analysis, auditing, methodological design and Python development
**Status:** five-source technical pilot completed · methodological review pending
**Last updated:** 14 August 2026
**Technologies:** Python · pandas · Jupyter · Power BI · DAX · Git · CSV/Parquet · APIs and open data
## The project in one sentence
I designed and developed a methodology to discover, audit, transform and integrate humanitarian and operational data sources for Niger without losing their granularity, traceability or meaning.
This case demonstrates how I work at the intersection of **operations, data quality, analytics and Python development**: understand the real problem first, formulate explicit rules and automate only what can be validated.
## Current outcome
The pilot methodologically integrates five components—**Kobo, INFORM Risk, WFP Food Prices, INFORM Severity and Network Performance**—and is technically complete in the `develop` branch of the organisation's private repository.
| Completion evidence | Result |
|---|---:|
| Sources documented by layer | 5 |
| Reproducible notebooks | 7 (`00`–`06`) |
| Notebooks executed end to end | 7/7 |
| Execution errors | 0 |
| Cells without identifiers | 0 |
| Final `integration_analysis` tests | 15/15 |
| Tests validated across the project | 41/41 |
| Methodological commit merged into `develop` | `3115c13` |
Technical completion does not yet mean methodological approval by the supervisors or the existence of an operational early-warning system.
## Problem addressed
Data relevant to a potential anticipatory-action system is distributed across open-data portals, APIs, Excel files, Kobo surveys and Power BI semantic …