A national mapping agency cut title errors by 94% across 48 million parcels

The challenge

The agency maintained the authoritative digital cadastral database for an entire country — more than 48 million parcels accumulated over decades of surveying, digitisation, and successive migrations between systems. Every migration had left sediment: duplicate geometries, slivers between parcels that should have shared a boundary, and titles whose recorded area no longer matched their mapped extent. Front-line staff spent their days chasing these inconsistencies by hand, and the backlog only grew.

The trigger for change was a national land-titling reform that required a verified, consistent cadastre as its foundation. The agency could not reform titling on top of a database it did not trust.

What we built together

The Airfree Cadastre Engine became the system of record, ingesting the full 48-million-parcel dataset and running automated topological validation across every geometry: overlaps, gaps, unclosed rings, and area-versus-title mismatches were flagged systematically rather than discovered incidentally. On top of that, the AI reconciliation pipeline matched satellite-detected boundaries against the DCDB to surface parcels that were occupied but unregistered.

• Full ingestion of 48M+ parcels into the Cadastre Engine as the system of record.
• Automated topology validation flagging overlaps, gaps, and area mismatches at scale.
• AI reconciliation surfacing occupied-but-unregistered parcels for surveyor review.
• Every correction written back with full provenance and an audit trail.

How the rollout ran

Rather than a big-bang cutover, validation ran district by district so staff could work through flagged errors in manageable batches. Critically, no correction was applied autonomously — the engine proposed, a cadastral officer confirmed, and the change was recorded with who made it and why. That kept the reform legally defensible, which for a land registry is not optional.

The result

In six months the agency processed the full dataset and reduced title errors by 94%, achieving full DCDB integration with automated boundary validation now running continuously rather than as a one-off clean-up. The verified cadastre became the trustworthy foundation the national titling reform had needed, and the validation that once consumed staff time now runs in the background, catching new inconsistencies as they arise.