How we clean DLD data
Two portals can both say "DLD data" and print different numbers. The difference is always the cleaning. Here is ours, in full, so you can judge it.
What goes in
Registered residential sale transactions from the DLD open data platform and registered rental contracts from Ejari, aggregated over rolling 12-month windows. No listings, no asking prices, no broker submissions, no estimates. Current window: 12 months to 2026-07-24, 177,749 sales across 44 qualifying areas.
What gets filtered
- Areas below minimum sample thresholds are excluded from rankings rather than estimated.
- Per-square-foot figures are suppressed wherever fewer than 10 registered sales back them.
- Metrics that fail sanity gates (yields outside 1 to 15%, growth outside -40 to +60%) are shown as null, not published. These gates exist because a handful of mis-keyed registrations can bend a small area's median.
- Non-residential and bulk-transfer records are excluded where identifiable.
- Project name variants in the registry are normalized before per-project stats are computed.
Why our numbers differ from other portals
- Portals showing "average prices" from their own listings measure asking prices; we measure registered sales, which usually sit below asking.
- Some data sites print raw feeds without junk-row filtering, so obvious mis-registrations (a full apartment at AED 50,000, a 30 AED/ft² sale) flow into their averages. Ours are gated.
- Averages and medians differ by design: we use medians because a single palace sale should not move a whole area's number.
- Window boundaries differ: our figures state their window and refresh date on every page.
Corrections
When we get something wrong, it goes on the public corrections log. If a number here disagrees with another source and you want to know why, message us; that question is our favourite kind.