What a basement flag should tell you about itself.
A boolean is not a decision. A basement flag that reports its own provenance and error rate lets you route the doubtful cases instead of rejecting them.
Basement is published as two fields - declared from inspection and record evidence, and predicted where modelled - so a rule can weight the evidence rather than act on a silent blend.
A basement changes a valuation, a construction risk assessment and a flood exposure all at once. So lenders write rules about it. The trouble is that most basement data is a bare true or false, and a bare boolean cannot tell you the one thing you need: how much to trust it.
Two fields, on purpose
We publish the basement as two separate fields. The declared field carries inspection and record evidence. The predicted field is modelled and tested against the declared set. They are never blended into one ambiguous flag.
That separation is what lets a rule behave. A modelled basement with no declared confirmation is not the same case as a basement confirmed on the record, and your routing should not treat them as if they were.
IF basement_predicted = TRUE AND basement_declared = NULL -> route to desktop before AVMNo AVM where a basement is suspected costs less when the flag states its own error rate.
The over-decline problem
A blanket rule that pulls every suspected basement off the AVM path is expensive. It routes cases to a desktop or a doorstep that did not need one, and it does it because the flag it is acting on refuses to say how often it is wrong.
A flag that publishes its tested accuracy changes the economics. Now the rule can send the genuinely uncertain cases for a closer look and let the confident ones through. The error rate lives in the release note, per period, on request - not in a headline.
What sharpens the decision
- -Storey count, as a cross-check on unusual configurations.
- -Floor area, which a basement usually moves.
- -Planning history, which records the conversions a fact find never sees.
A basement flag should tell you three things: whether there is a basement, where that claim came from, and how often it is right. Two of those are missing from a bare boolean. That is why we ship all three.