Valuation infrastructure
Price real assets on real evidence.
A Bayesian pricing engine that reads a property document, locates it against recorded sales, and returns a defensible valuation — with its sources and a model hash attached.
1.3M+
Real recorded sales in the Price Graph (French land registry, DVF)
36k
Communes locatable by address — nationwide geocoding, not Paris-only
6.9%
Median absolute error, back-tested on held-out sales
10k paths
Seeded Monte Carlo · P10 / P50 / P90 · reproducible
01
Bring the asset
Upload a listing, cadastre, lease or teaser. The engine extracts the facts and geocodes the address to its real commune.
02
Price on evidence
A 5-layer Bayesian model anchors on recorded sales nearby, then a Monte Carlo band and SFDR class are computed.
03
Sign or flag
Measured inputs are labelled and unknowns are shown as assumptions — the attestation is withheld until the asset is grounded.
Under the hood: document → extraction → geocode → recorded-sales prior → Bayesian update → Monte Carlo → SFDR. Every figure carries its source.