How we get to a number, and when we refuse to.

Two registers do most of the work. The Dubai Land Department publishes every registered property transaction, and the Ejari system registers every tenancy contract. Both are records of deals that actually completed, which is the distinction that matters: an asking price is an opinion, a registered transaction is an outcome.

We read them through the Land Department’s open data channels rather than scraping listing sites, and we reconcile both against a curated building register so that a tower is the same entity in the sales file, the lease file and on the page you are reading.

A raw transaction file is not a market. It contains gifts between family members, bulk portfolio transfers, corrections, and rows whose price and area cannot both be true. Left in, they drag averages around and make a quiet building look volatile.

Before anything is calculated we drop non-sale transaction types, anything under AED 50,000, and rows whose price per square foot falls below a floor that no genuine arm’s-length sale in Dubai reaches. We never mix property types: land, villas, apartments and hotel units are separate markets and averaging across them produces a figure that describes none of them. Ready and off-plan stay separate for the same reason.

Medians do the reporting rather than means, because one penthouse moves a mean and tells you nothing about the two-bedroom you own.

The most common failure in property data is a number computed from too little evidence and presented with the same confidence as one computed from thousands of records. We would rather be visibly silent.

A building only gets a published figure when it has enough registered sales and enough tenancy contracts in the recent window to support one, with at least one bedroom cohort deep enough to stand on its own, and evidence that the market there is still live rather than historic. Applied to our own building register, that standard admits roughly a fifth of the buildings we hold data for. The rest resolve to their community, where the sample is real.

Where a yield is shown it is matched: the rent and the price come from the same building, the same property type and the same bedroom count over an equivalent period. A rent drawn from one population divided by a price drawn from another is arithmetic, not a yield.

Four jobs, none of them on screen. Matching: one tower can appear under half a dozen spellings across the sales file, the lease register and the owners-association records, and a model reconciles them so your building is compared with itself. Cleaning: gifts, bulk transfers and related-party sales look like ordinary transactions until they are set beside the building around them, and a learned filter catches what a price floor misses. Weighing: among the sales that survive, the ones that resemble your unit in size, floor, age and layout count for more than the ones that do not. Reading: photograph a title deed, an Ejari certificate or a Mollak invoice and a model returns the fields themselves, and returns nothing where a figure is not on the page rather than infer one. What happens to that document afterwards is set out in the privacy policy.

None of this is a chat box, and none of it is a forecast. A model here decides which records belong together and how much each one should count. The arithmetic that turns them into a figure is set out above, and you can follow it line by line.

Nothing here predicts a price, and nothing here recommends a sale. The same standard applies to the models as to everything else on this page. Where the evidence behind a figure is thin, no amount of modelling earns it a number, and we publish nothing. As the models change we will say so here, with the date and what changed.

Service charges are not current.The owners-association register we can access has not been updated upstream since January 2024. A service charge question means “what am I paying now”, and a 2023 budget year does not answer it, so we do not put that figure forward as though it did. Your own rate is visible in Mollak through the Dubai REST app, and our guide explains how it is set.

Our estimates are not formal valuations.They are computed from comparable registered sales. When a bank, a court or a counterparty needs a number that carries weight, that is a RERA-registered valuer’s job, and the valuation guide covers what they do differently.

Nothing here is legal or financial advice.Where we describe a rule, we name the instrument — Decree 43 of 2013 for rent increases, Law 33 of 2008 for notice periods — so you can check it rather than take our word for it.

Registered records get amended, buildings get renamed, and matching a commercial tower name to a Land Department record is not always unambiguous. We hold ourselves to publishing only exact entity matches, and where two buildings would end up sharing a name we drop the page rather than risk attaching one tower’s prices to another.

If a figure on this site looks wrong for a property you own, tell us and we will check it against the record and correct it. Write to us with the building and we will come back to you.