How we compare
A number is only useful if it's compared to the right thing
Dubai property is full of true numbers arranged into misleading comparisons. A launch price beside a market average. A yield built from one building's rent and another's price. A “15% growth” that counts a different set of homes at each end.
Keyva is built on 1,353,718 Dubai Land Department sale records and 3,459,821 registered Ejari tenancies. The data matters less than the discipline applied to it: compare like with like, and say nothing when you can't.
What that means in practice
A developer's “from” price against an area median
A starting price is the cheapest unit in a development. A median is the middle of what sold. Subtracting one from the other understates the real gap on a comparable home — so we show both figures and say they are not like-for-like, rather than printing a tidy premium percentage.
Both numbers, side by side, labelled by what they are
Rental yield from mismatched sources
DLD files sale records under community names like Dubai Marina, and tenancy records under official sectors like Marsa Dubai. Taken at face value, that produced 74 areas with sales and no rents — and no yield at all for most of the places people actually search.
The two are joined only where shared projects prove they describe the same place, and where a sector spans several communities, the page says the rent side is broader
Year-on-year change from a sample too thin to carry it
Keyva holds 20 years of monthly price history, so the comparison is available almost everywhere — but availability is not the same as meaning. An area with nine sales this quarter against six a year ago can produce any percentage you like, and the arithmetic will not tell you it is noise.
The comparison is published only where both ends clear the minimum sample, and the sample count is shown beside it
Rent trends from months we only partly captured
DLD's exports filter by registration date, so our first rent download captured old contracts only if they were registered late — Dubai-wide it read as 717 tenancies in January 2025 climbing to 68,862 a year later, explosive growth that was really our own coverage arriving. The full archive since loaded does not have that bias, but the guard it taught remains: a month far thinner than its neighbours is dropped from trends as partial coverage.
Those months are dropped from trends entirely, because a biased sample is not a smaller sample
Bulk leases counted as homes
One Ejari contract can cover forty units at a single price. Averaged in with individual tenancies, that quietly drags every median it touches. Roughly one in five of the contracts in the full Ejari archive is a bulk lease.
Excluded and counted, not divided and hoped over
A developer's average price read as a quality premium
A builder's raw price per square foot mostly measures which areas they build in. Estate Grand posts the highest headline figure of any developer in our records, AED 5,942 per square foot — but compared against the same areas, the same completion status and the same months, their homes sell about 6% above the local market. Atlantis The Palm 2, on a lower headline, commands 152%. Emaar Development, the largest by matched volume at 47,513 sales, is 23% above; Azizi and Binghatti, also among the largest, are within 3%.
Every developer measured only against sales in the same area with the same off-plan or ready status, with the raw figure shown beside it
Off-plan and ready stock averaged into one median
These are two different markets. Over the last twelve months of our own DLD records, off-plan trades about 27% above completed homes city-wide — AED 1,796 per square foot against AED 1,417. By area the gap is nothing like uniform: across the 51 areas carrying at least thirty registered sales on each side, the middle one shows 35%, and the range runs from roughly 7% to over 100%. Palm Jumeirah's off-plan stock records AED 5,534 per square foot against AED 2,635 for ready homes. Since off-plan is now the majority of Dubai sales, a blended median mostly tracks the off-plan side while appearing to describe the whole market.
Both medians shown separately on every area, with the gap between them stated
“Data to” a date read as “everything up to” that date
DLD's transaction export carries the date a sale happened, not the date it was registered with them. A sale agreed in the last few weeks may not be in the file yet, so the newest month is always partial and always grows on the next pull. We have no registration date at all, so we cannot tell you how partial — only that the most recent month is the least complete one on every page.
Dates are labelled as transaction dates, the newest month is treated as incomplete, and the year to date is never compared against a full year
Registrations that are not prices, counted as prices
The register is a record of filings, not a price list: partial payments, transfers between related parties and nominal considerations sit in the same column as real sales. An 807 sqft apartment logged at AED 26,130 is the shape of it. Every sale figure on this site is bounded to AED 200–20,000 per square foot and a total of at least AED 100,000. That is about 1% of rows, invisible in a median and decisive in a minimum.
One shared bound, applied in every query, so the same sale is never counted on one page and excluded on the next
A service charge quoted as this year's bill
Owners' association budgets are filed FOR a year, and the newest complete year in the record is 2023. They are matched to a project by name, which is the only key the filings carry. So the figure answers what a building budgeted two years ago, not what an owner will be invoiced this year.
Labelled with the year it was filed for, and never presented as a current charge
Land blended into home prices
A plot of raw land and a two-bedroom apartment are both DLD transactions. Averaged together they produced area medians as low as AED 5 per square foot, which is not the price of anything a person can live in.
Residential stock only, on both the sale and rent sides
Names in Arabic and English
You can search this site in either script. Type مرسى دبي or “Dubai Marina” and you reach the same page.
That works because the Land Department publishes most names twice, once in each script, on the same row of its own files — the Arabic name of an area beside the English one, and the same for projects, towers and developers. We lift those pairs straight out of the register, so the Arabic name on a page is the name DLD registered, not a translation we produced. Of the 3,039 registered projects, 2,968 carry an English name somewhere in DLD’s files.
A search also forgives the ways Arabic is spelled differently without changing the word: the hamza forms, the two endings that sound alike, short vowels, and the definite article. It never transliterates. Typing Latin letters will not match an Arabic name by sounding it out — the two scripts meet only through a pair DLD actually published, because guessing one would put a name on a page that no register contains.
The floors, and where each one is enforced
Below these counts a figure is withheld rather than printed. They are read from the code that enforces them, so this table cannot drift from what the site actually does.
| What is being shown | Minimum | Counted in |
|---|---|---|
| An area gets a page in search | 10 | registered residential sales |
| A project gets a page in search | 5 | registered sales |
| A building gets a page at all | 20 | registered home sales |
| A listed unit's like-for-like sales | 5 | sales in that building with the same bedroom count and a size within 10%, last 24 months |
| A rental district gets a page | 200 | tenancy contracts, last 12 months |
| A month appears in a price series | 5 | sales in that month |
| An area's gross yield | 10 | ready sales, last 12 months |
| Each side of an area's ready / off-plan split | 20 | sales on that side |
| A ready or off-plan median on /compare and in the market report | 60 | sales in three complete months |
| A project's resale or off-plan median on /compare | 20 | registrations in twelve complete months |
| A quarter in the prices-by-quarter tables | 60 | sales in that calendar quarter |
| Any rent median | 20 | registered tenancies |
| A bedroom row | 15 | sales, all years |
| A bedroom row's recent price | 10 | sales, last 12 months |
| A row in the yields table | 500 | ready sales, last 12 months |
| …and its rent side | 1,000 | registered tenancies |
| A price-band share | 100 | sales on each side |
| A villa median | 50 | registered sales |
| A developer's premium | 30 | sales in one area |
| …against the same area, in each month compared | 15 | sales that month |
| …and a developer total of | 150 | matched sales |
Two rules sit behind the whole table. A median is not an index: it is the middle of what sold that month, and it moves when the mix of what sold moves. And ready and off-plan homes are never pooled into one figure — they are two markets, and over the last twelve months off-plan trades about 27% above completed stock city-wide.
Why we would rather show nothing
A blank where a number should be looks like a worse product. It is a better one. Every figure Keyva declines to show is a figure a competitor is prepared to guess at, and a buyer is entitled to assume both were measured the same way. When something here is missing, the page says why — and usually says what would have to be true for it to appear.