i · Will not
Widen the property type to reach a number.
A flat is compared to flats. If the cohort is thin, the methodology widens geography before it ever widens type — and discloses the stage it stopped at.
Methodology
Every claim has a name and a source.
The value is not in any single dataset — anyone can look up a sold price. It is in the intersections: what the registered sale record can say once it is matched to a property's structural attributes, corrected by the house-price index, and joined to its official geography. An estimate that leans on one source produces a guess. A methodology that integrates sixteen produces intelligence.
Same market, two methods
Why an algorithm's guess isn't evidence.
How algorithms guess
-
Lagging data
Most portal estimates rebase against an index that updates monthly or quarterly — a model of last cycle's market, applied to this week's question.
-
Anchored to the asking price
Many valuation models are trained on, or nudged toward, what the property is listed at — which is the seller's opening position, not evidence of what it's worth.
-
Every sale weighted equally
A family transfer at half price and a genuine sale are averaged together, and the model never says which is which.
The SettleIQ standard
-
HM Land Registry, directly
Every figure traces to a registered, completed transaction — the price that was actually paid, not the price that was asked.
-
Floor area, cross-checked
The floor area behind every anchor is taken from the floor plan or listing and checked against the EPC register. Where the two disagree by more than 10%, the folio says so before it gives you a number.
-
Extreme tails trimmed
The lowest and highest £/sq ft in every cohort are set aside before the median is taken. The bottom of the range is where non-market transfers, probate and distressed sales concentrate; the top is where full renovations land. The trimmed sales are still listed in the folio, marked, so nothing disappears silently.
From raw registers to a number
The filtration waterfall.
Four stages, always in this order. Nothing reaches the tactical bracket without surviving all four.
-
01
Data ingestion
HM Land Registry Price Paid Data and fifteen other public registers are pulled in raw — nothing pre-filtered, nothing pre-modelled.
-
02
Tail trimming
Once the cohort is assembled, the lowest and highest 5% by £/sq ft are set aside before anything is averaged. They stay in the folio, with the reason, so the number can be checked against them.
-
03
Square footage standardisation
Every surviving sale's floor area is cross-referenced against the EPC register, so price-per-square-foot is compared on a registered basis, not a listing agent's estimate.
-
04
The tactical bracket
The cohort's standardised £/sq ft sets the Glass Box Anchor. Four leverage signals then position the Open, Aim and Walk prices around it.
The rule that does the work
The Identity Lock
Identical is not a vague word here. The methodology locks six dimensions before a single transaction is allowed into your cohort. A two-bed flat is matched only against other two-bed flats.
- Property typeFlat, terrace, semi, detached. Never blended to reach a larger sample.
- TenureFreehold and leasehold are different legal assets and are never compared to each other.
- Age bandConstruction era shapes a property's cost base and its likely buyer — an older conversion and a modern block sit in different markets, even streets apart.
- Floor bandBedroom count and floor area are read together, so properties of meaningfully different size are never read against each other.
- New buildA new-build premium is real and it does not survive resale, so new builds are held as their own class rather than blended with second-hand stock.
- Plot sizeFor detached and semi-detached houses, the land itself is part of what's being priced, so it's weighed alongside the structure rather than ignored.
The evidence base
Sixteen datasets, in the order the methodology consumes them.
Identity first, then price, then time, then geography, then liability, then amenity. The numbering below is the processing order, not a ranking.
-
01 VOACouncil Tax Valuation List Valuation Office Agency
Identifies the Council Tax band for the property in question.
Verify -
02 EPCEnergy Performance of Buildings Register MHCLG
Property attributes and the floor area used by the anchor — cross-checked against the operator-supplied figure, with any disagreement above 10% stated in the folio.
Verify -
03 HMLRPrice Paid Data HM Land Registry · 1995 — present
Every registered transaction behind the cohort. Named, dated, and verifiable at landregistry.data.gov.uk.
Verify -
04 UKHPIUK House Price Index ONS / HM Land Registry
Time-adjusts historical sales to today at local-authority-month resolution. A 2023 sale is not a 2026 comparable until it is corrected.
Verify -
05 PIPRPrice Index of Private Rents ONS · official statistics since May 2026
Rent levels by local authority, property type and bedroom count. The comparator behind the five-year cost of ownership — what renting the same home would have cost instead.
Verify -
06 NSPLNational Statistics Postcode Lookup ONS
Official geography. Defines the postcode sector the identity lock searches first, and the cascade it widens into only if it must.
Verify -
07 NaFRANational Flood Risk Assessment Environment Agency · NaFRA 2025
Flood exposure from rivers, sea and surface water. Feeds the insurance-risk severity scale and the conveyancing question list.
Verify -
08 RadonRadon Potential Classification BGS / UKHSA
Whether the address sits in an affected area and at what class — a remediation cost that is never in the listing.
Verify -
09 NoiseStrategic noise mapping Defra · Round 4
Airport-noise contours at postcode level, in Lden. Whether the address sits inside a band that affects insurability or resale — a cost that is never in the listing.
Verify -
10 PlanningPlanning Data (conservation areas, Article 4) MHCLG
Whether permitted-development rights have been withdrawn. Determines what a buyer can and cannot change after completion.
Verify -
11 NHLENational Heritage List for England Historic England
Listed status on the building or its immediate context, with the consent regime that follows from it.
Verify -
12 IMDIndices of Multiple Deprivation MHCLG · 2019 + 2025
Living environment, income, crime and barriers domains at LSOA level — read as directional context, never as a verdict on a street.
Verify -
13 GIASGet Information About Schools + Ofsted DfE / Ofsted
Catchment-relevant school records and current inspection grades, distance-ranked from the address.
Verify -
14 NaPTANNational Public Transport Access Nodes DfT
Every stop, station and interchange within walking range — the transport claim in the listing, checked.
Verify -
15 OfcomCoverage API Ofcom
Fixed-line availability at the premises, from a live per-address query. Advertised speed is a marketing figure; this is the record.
Verify -
16 OSMOpenStreetMap OpenStreetMap contributors
Amenity proximity and the physical context around the plot — what is actually next door.
Verify
The constraints
Three things this methodology will not do.
Each of these makes the product harder to build and occasionally makes it return nothing at all. Each is also the reason a folio can be argued from in a room with a professional negotiator.
ii · Will not
Rest an anchor on a floor area it hasn’t checked.
The listing’s square footage is a claim. A 5% measurement error on a £647/sq ft flat is roughly £24,000 — several times the width of the offer bracket.
iii · Will not
Name a number it cannot defend.
The search widens geography, then time, then the era band — never the property type — and names the stage it stopped at. If fewer than five registered transactions remain even then, the methodology returns no bracket and no invoice. Silence is sometimes the honest output.