Skip to content
YIELDMAP

Report

What this project set out to do, what it measured, and what it does not know.

The problem

Dubai’s property registry is public, unit-level and daily. Every sale and every registered tenancy is recorded. Yet buyers, tenants and small investors still decide from broker claims and portal listings, because nothing turns that registry into a valuation, a yield, a risk score and a plain-language explanation that can be checked.

YIELDMAP does that, and its distinguishing property is that every number it shows can be traced back to the query that produced it. That is not a promise in a document — a figure cannot leave the API without carrying its own SQL, and a test at the boundary fails if one does.

What was measured

  • Valuation. not yet computed.
  • Forecast. not yet computed.
  • Retrieval. not yet computed.
  • Cost. Zero. Every model backend in the chain is a free tier, the paid one is off behind two switches, and requests are counted in a durable ledger.

What it does not know

  • · Service charges are an estimate, and they are the largest input to any net yield here.
  • · Vacancy cannot be measured from the open data at all, so it is assumed.
  • · Lending caps are carried as assumptions because the Central Bank site refuses automated clients.
  • · Yields use medians for a community and unit type, not the rent of a specific unit.
  • · Community shapes are curated centroids rendered as hex cells, not boundaries.
  • · Forecasts are projections under stated assumptions. Nothing here is advice.

Read further