Asset intelligence infrastructure for Uzbekistan
ASINT — automated valuation, location intelligence and risk on one data layer. We are not building a price calculator; we are building the layer that decisions about property, land and vehicles are made on.
"What is it worth" is only one of three questions. A bank, an insurer, a developer and a government need two more: "what is this place" and "what can go wrong here". We answer all three.
Why this did not exist before
Uzbekistan is a closed-transaction market. Transaction prices are almost impossible to find in open sources. Everything else follows from that.
There is no MLS. There is no public price index built on actual transactions. A bank has no instrument to verify what collateral is worth, and an appraiser works from classified ads and personal experience. A manual appraisal takes 3–7 days and costs $200–500 per object, and overvaluation of collateral by 15–30% is a known feature of the market.
Meanwhile the market is large and growing: according to the Central Bank of Uzbekistan, 2025 saw 319,500 property transactions (+15.8% year on year) and 21.2 trillion UZS of mortgage lending (+29%). Roughly 22% of deals are mortgage-backed — about 70,000 objects a year that require a collateral valuation.
We solved this differently. In place of a transaction registry that does not exist, we run our own listing-data pipeline with deduplication, outlier removal and listing-quality scoring; hedonic regression to separate price movement from changes in the composition of supply; and gradient boosting over three years of history. The same hedonic methodology is used in quarterly reports by the Central Bank itself, and by leading automated-valuation operators in developed markets — so this is an established method, not an invention of ours.
Three layers, not four products
It is easy to read this site as four equal products, one of which happens to be a map. That is not the structure:
Applications
What is it worth?Link Home, Link Auto, Link Land — valuation of apartments, houses, land and vehicles. Collateral portfolio monitoring for banks.
Intelligence layer
What is this place?GeoScore: three axes, nine subindexes, a 0–100 score for every cell in the country. Plus momentum — where a place is heading.
Foundation
What does all of it stand on?A registry of 329 indicators with an absolute-scale methodology, and a national H3 hexagonal grid: 14 regions, 206 districts, 9,431 mahallas.
Which leads to the point that matters: GeoScore is not a sibling of Link Home in a product line-up — it is the layer underneath it. Every object in our database receives a GeoScore the moment it is ingested, and the valuation model consumes it as a feature. That is also why price is forbidden inside GeoScore: including it would make the index circular.
Why 50 is the norm, not the middle
An ordinary location rating is a percentile: "70" means "better than 70% of the rest". Such a score moves when other places move, and it does not travel between cities. We made the scale absolute: for every indicator the 0 and 100 points are taken from a source outside the project — Uzbekistan's urban planning norms (ShNQ), sanitary norms (SanEpid), WHO guidelines on air and noise, the 3-30-300 greenery rule, the DEGURBA settlement classification, World Bank seismic hazard maps.
What this changes in practice
A pair of values like "Jizzakh 34, Tashkent 80" no longer reads as "Jizzakh is in the bottom third". It reads as "the average location in Jizzakh falls short of the norm; the average location in Tashkent exceeds it". That is a different statement: it describes a gap against a standard, not a position in a queue.
- A place's score does not change because some other place changed.
- Two points at opposite ends of the country are directly comparable.
- The gap to the norm can be counted in money and in objects — which makes it a planning instrument rather than a ranking.
The numbers
Explainable, not a black box
Every valuation returns the top five factors that produced it, a confidence interval and a confidence level. A valuation you cannot defend to a credit committee is a valuation you cannot use.
Published methodology
We publish the structure of the index, the data sources and the external anchors of the scale. That is what separates a data provider from a lead-capture form.
Honest coverage
We name the indicators we do not yet have and the subindexes that currently ship in reduced form.
Speed as a consequence
A result in seconds where it used to take days. But speed on its own is worth nothing: it only means something together with an interval and an explanation.
The methodology is open
How GeoScore is built, which data sources it uses and where the anchors of its scale come from are described publicly. If you are evaluating us before an integration, start there.
See GeoScoreWant to go through the methodology?
We will walk you through how the models and the index work, and what applies to your problem
Contact us