Three years on, jittering is still our default approach for working with sensitive data. The original problem hasn't gone away: how to de-identify private health information without destroying the contextual signal our machine learning models depend on.

Traditional redaction strips out too much. Jittering adds controlled statistical noise instead, hiding individuals while keeping the underlying patterns intact. Since that first project, we've reused the technique on plenty of other datasets.

We have four interactive mapping applications live, all built around geospatial data with an environmental message, encouraging a reduction in carbon fuel usage, healthier active travel choices and promoting a host of green initiatives.

Each project uses a combination of the Leaflet JS library, Open Street Map tiles and GeoJSON, managed by our flat file content generator, to provide a rich user experience browsing locations & activities and choosing routes across the South West of the UK.