Valuation signals and underwriting

Fabrica deliberately surfaces more than one valuation signal for each property and remains open to additional valuation systems over time. No single model is treated as ground truth. The value of multiple signals is in their agreement and disagreement: independent estimates that converge raise confidence, while estimates that diverge are a signal to be cautious.

The signals

• FabricaAVM. Fabrica's in-house model for raw land, with a confidence level per estimate.
• Prycd. A third-party land valuation model. Fabrica publishes the Prycd signal alongside its own on property pages, so users can compare two independent estimates side by side. Keeping a parallel third-party signal visible reflects a deliberate openness to multiple valuation systems rather than a single proprietary number.
• Assessed value. The county tax assessor's value. Useful for tax context, but typically well below market and updated on irregular schedules.
• Last recorded sale. What a buyer actually paid, where a recent arms-length sale exists.
• Lending pool valuation. For pool-based lending, the pool applies its own valuation and risk filters on top of the available signals (see Pool-Based Lending).

How lending uses them

Lending does not act on a single estimate. An underwriting view triangulates the available signals and sets conservative loan-to-value accordingly:

• When independent signals agree, confidence is higher and terms can be set with a normal margin.
• When signals diverge, or when confidence is low (for example a remote parcel with few comparables), the system stays conservative, widening the margin or declining.

This is intentional. Several imperfect, independent estimates that can disagree are safer to underwrite against than one estimate trusted blindly. It is also why Fabrica keeps a third-party signal in view next to its own.

The lending venues consume these signals differently:

• Pool-Based Lending (pool-based, the primary live integration) uses estimated value as a starting point and applies additional risk filters to determine automated loan terms.
• Peer-to-peer lenders evaluate properties individually and set their own terms; the signals inform but do not determine their offers.

For capital providers and researchers

Because the signals are published on property pages, written into token metadata, and available as a signed on-chain reference price (see Price Oracle), capital providers can build their own underwriting policies on top of them, and researchers can study land-valuation behavior across a real, growing set of tokenized parcels. Live protocol and lending statistics are available on the Dune dashboard.