The working, written down.
Notes on valuation, confidence, verification and provenance - for the people who route lending cases on data, and the model-risk teams who have to sign off on how.
The confidence level is the product.
A value without a confidence level is a liability wearing a decimal point. The number your routing rules actually run on is the one next to it.
Sold prices are the only honest benchmark.
A sale-value estimate should be evaluated against completed sales, with the cohort, date, metric and limitations published together.
The second value: test before changing a decline route.
A second valuation can be evaluated as another governed signal where the first view is uncertain. Recovery is a hypothesis, not a promised outcome.
Routing rules that survive model-risk review.
A rule that acts on a silently blended field cannot be defended. Publish which fields are declared, which are modelled, and how the models test.
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.
High-value stock and the pull to the mean.
Above a million pounds, sparse comparables can pull an estimate towards the area average - exactly where the exposure is largest.
Fact finds lie politely: verification patterns.
An applicant is not dishonest. They just do not know the floor area, forgot the loft conversion, and rounded the bedrooms up. Verify the record, do not trust the recall.
The back-book exercise, step by step.
A UPRN list in, a flat file back. Revaluation context, attribute refresh and physical-risk views, starting with a bounded evaluation.
EPC deltas as a product input.
The current rating is disclosure. The gap between current and potential is a product - the retrofit headroom a green tranche is designed around.
Property risk for lenders: evidence and limitations.
Address-level flood and ground indicators can support origination, collateral review and portfolio monitoring. Here is what the evidence proves, what it does not and where people stay in the loop.
Change logs, additive columns, calm releases.
A data supplier that breaks your schema on a Tuesday is a supplier your model-risk team will never forgive. Old fields keep working.
Northern Ireland for lenders: what's actually there.
Attributes and rebuild at near-Great-Britain depth. No valuations, no sold prices, no planning yet. We tell you which layers are there and which are not.
Bring the cases these notes describe.
Choose a controlled sample or API trial for the workflow these notes describe, with the provenance attached.