Why valuation models keep missing transformed houses

Most valuation models see a property as a collection of current attributes. Floor area, bedrooms, property type and location.

What they often fail to infer is the transition between two states.
A detached house may now be 267m², but its previous records describe a 168m² property. Its last sale may predate two extensions. Most nearby comparables may still be smaller, unaltered homes.

Viewed separately, each piece of evidence looks incomplete. Taken together, they show that the property has become a materially different asset.

This distinction mattered in our testing. Detached houses with verified evidence of substantial physical transformation were underpredicted by approximately 2% on average against eventual sale prices.

Simply being large or unusual was not enough. Applying a general premium to locally exceptional houses produced the wrong result and sometimes increased overvaluation. The relevant signal was not size alone, but the relationship between the property’s previous and current state.

That is the inference conventional models often miss. Rare properties are pulled back towards the local norm, even when there is credible evidence that they no longer belong to it.

Not every extension adds value. Layout, build quality and lost garden space still matter. But the valuation should at least recognise that it is assessing a different asset.

Property data is usually treated as a snapshot. For altered homes, the history of how the property changed can be just as important as its current measurements.