Case Studies
End-to-end reproductions of real DataKind UK
charity projects, rebuilt with KindTech. Each one chains several connectors on a
shared geography_code, and — because real client data is private — pairs the
public data with a transparent synthetic stand-in so the full analysis runs.
Every study has a runnable marimo notebook: open it in the
molab cloud runtime to run it in your browser (it fetches live ONS data), or
locally with uv run marimo edit examples/<notebook>.py.
| Case study | What it shows | Run |
|---|---|---|
| Smart Works | Women's unemployment vs service reach (Census 2021 + LAD boundaries) |  |
| Citizens Advice Lewisham | Deprivation vs service usage (postcodes + IMD + LSOA boundaries) | |
| Sobus | BAME mental-health referrals in Hammersmith & Fulham (outcodes + Census ethnicity) | |
| Starlight | Hospital play provision vs need across ICBs | |
| Material Focus | Travel time to the nearest recycling point (LAD boundaries + population) | |
How each study is structured
- Problem statement — the question the charity set out to answer.
- Dataset involved — the open (and, where relevant, synthetic) data used.
- Desired output — what the original analysis produced.
- Reproduction with KindTech — the connector pipeline, with figures.
- Lessons learned — takeaways and where the approach generalises.
Have a case study to share?
Using KindTech in your own work? We'd love to feature it — open an issue or PR on the GitHub repository.