The Trust Ledger
We publish our working — and our mistakes
A platform that asks you to trust a single number has to show how that number is made and own it when it gets something wrong. This is the append-only record of every change we make to the Intelligence Score methodology and every correction we issue — dated, with the reason. Entries are never edited or removed; if a past change is reversed, the reversal is logged as a new entry.
The proprietary weights are not published — but every dimension that feeds the score is named in the methodology, and every change to how those dimensions are computed lands here.
- Methodology change
Split off-plan / ready benchmarking (median-relative)
The Intelligence Score previously read every unit against a single secondary-market price-per-sqft benchmark per district. With roughly 93% of the live feed being off-plan, that systematically penalised off-plan units against a ready-market reference and pushed almost everything to the WATCH band. The model now benchmarks each unit median-relative within its own cohort (off-plan vs ready) per district, so an off-plan unit is judged against off-plan peers and a ready unit against ready peers. Effect: the label distribution now reflects genuine relative value rather than a structural off-plan penalty.
- Coverage
District benchmark coverage expanded
Added internal price-per-sqft and yield benchmarks for additional Dubai districts so fewer listings fall back to a generic city-wide reference. Wider coverage means a unit is more often scored against its own neighbourhood rather than a blunt average.
- Correction
Building-DNA false-positive on a service-charge alert removed
An automated building-data check flagged a service-charge year-over-year movement that, on review, was a data artefact rather than a real increase. The alert was withdrawn and the check tightened to require a corroborated multi-period trend before it surfaces a warning.