How we price and forecast

Short version: we start from deals that were actually registered, we show you the deals, we say how sure we are, and we keep score in public.

Where the numbers come from

Sale prices and rents come from open data published by the Dubai Land Department (DLD) and Dubai Pulse: registered sales and registered leases (Ejari). Each community page shows its sources and the date the data was retrieved.

We do not use asking prices from listing portals as evidence. An asking price is what someone hopes to get; a registered deal is what someone paid.

Price ranges

By home type, not by community average. A 3-bedroom original villa and a 5-bedroom villa on a large plot are different markets, so we group deals by type, size and, where the data allows, cluster.
Comparables you can see. Each range lists the recorded deals behind it, with date, plot, built-up area and price.
Ready and off-plan apart. Off-plan prices include payment-plan terms and delivery risk; mixing them with ready homes distorts both.
No single price per square foot for villas. Plot size, built-up area and condition (original or renovated) move villa prices more than total area alone, so we show them separately.
Not a valuation. A range from recent deals is a starting point for a conversation, not a formal valuation by a RERA-certified valuer, and it does not replace one for a mortgage or a legal process.

When we abstain

When a home type has too few recorded deals in the period, we do not publish a range or a forecast for it. The page says so and gives the count. A confident number built on two or three deals would be worse than no number.

Forecasts

Intervals, not a single number. We publish a central case (p50) and an 80% interval (p10 to p90): eight times out of ten we expect the outcome inside it.
Always next to a simple benchmark. Every model is compared with the simplest honest alternative: carrying today's median price forward.
Granular only where the sample allows. By community first, by cluster or home type when there are enough deals.

Track record

We test the model on past periods it never saw: we make the forecast using only data available at the time, wait the full horizon, and compare it with what happened. The community pages report the average error of our model and of the simple median, and how many predictions were checked.

If our model does not beat the simple median, we say so on the page. Every new forecast is stored with its date so it can be scored when its horizon ends.

What we don't do

We don't republish listings or content from property portals.
We don't advertise units for sale or rent. If brokers publish a unit in the future, it must show a valid Trakheesi (Dubai) or Madmoun (Abu Dhabi) permit, or it does not go up.
We don't let anyone pay to change a price, a forecast or a finding.

Found an error? Write to hello@fairsqft.com. We correct it and note the date of the change.