Evidence Boundaries in UK Property Intelligence
How evidence scope, identity, provenance, freshness, and decision eligibility make a property analysis easier to trust and challenge.
- Published
- Updated
- Author
- Tellingly Research
- Reviewed by
- Tellingly Product Governance
Introduction
A property lead can look convincing long before the evidence is strong enough to support a decision. An address may resemble a company filing, a local price benchmark may appear relevant, a planning event may sit nearby, or an energy certificate may describe a similar premise. Each observation can help an analyst, but they do not all establish the same thing about the selected property.
Evidence boundaries are the rules that keep those differences visible. They describe which subject a record belongs to, how it was matched, whether it is direct or contextual, when it was observed, where it applies, and whether it is eligible to influence a conclusion. They allow useful information to remain available without silently turning a nearby or candidate record into a property fact.
Executive summary
A defensible property-intelligence workflow begins with subject identity. Exact UPRN, title number and canonical-address evidence provide stronger anchors than a postcode, building name or geographic proximity. Once the subject is established, each observation should retain its source, scope, effective date, observation date, geography, match method, confidence, freshness and usage policy.
The presentation should distinguish at least direct, analyst-confirmed, candidate, building-level, contextual, historical, stale, conflicting, excluded and unresolved states. Missing data is not the same as a source failure; stale data is not the same as no data; and candidate evidence is not a zero-confidence fact. These distinctions determine what can support a headline conclusion, a scenario assumption or a report.
1. Establish the subject before interpreting the evidence
Reliable analysis starts by answering a simple question: which legal or physical asset is being discussed? A UPRN is a unique reference for a spatial address, while a title number identifies registered land. VOA and Energy Performance Certificate records use their own identifiers and address structures. These records can often be reconciled, but they are not interchangeable.
Address matching should preserve premise numbers, ranges, suffixes, flats, suites, units, building names, streets and postcodes. The first number in an address is not sufficient proof. Multi-let buildings, renamed sites, corner properties and address aliases make a superficially close match risky.
Where an exact identifier is unavailable, the system may present ranked candidates. The best candidate confidence should be shown separately from confirmed identity confidence, together with the reasons for the match and any conflicts. An analyst confirmation can create a durable reviewed link, but it should remain visibly distinct from authoritative source confirmation.
2. Classify evidence by scope
An observation's scope answers what it can safely describe. The same source family may produce several scopes depending on the match.
| Evidence state | What it means | Permitted use |
|---|---|---|
| Direct | The source row is linked to the selected property through an eligible identifier or approved exact match | May support a property conclusion when current, displayable and relevant |
| Analyst-confirmed | An authorised reviewer selected a source-backed candidate and recorded the reason | May support the reviewed property while preserving the underlying source confidence and audit trail |
| Candidate | A plausible property or party match requires verification | May be shown as a lead; must not silently affect a headline fact or score |
| Building-level | The record applies to a containing building or site rather than necessarily to the selected unit | May inform the building context with an explicit scope label |
| Contextual | The record describes a postcode, local authority, market, sector or comparison population | May support benchmarks and scenarios; must not become a direct property fact |
| Historical or stale | The observation is no longer current but remains relevant to change or provenance | May inform a timeline with its date and freshness visible |
| Conflicting | Eligible sources contain materially different values or relationships | Both positions should remain visible until reconciled |
| Excluded | The record is ineligible because of scope, licensing, quality, duplication or an identified mismatch | Retained in the audit where permitted, but cannot influence the decision |
| Unresolved | A decision-relevant question has no eligible answer linked yet | Becomes a named verification action rather than a reassuring substitute value |
3. Separate evidence confidence from evidence coverage
Evidence confidence asks how strongly eligible observations support a finding. Evidence coverage asks how much of the evidence required for that finding is present and current. A case can have strong evidence for one catalyst while still having poor coverage of ownership, occupation, planning or financial inputs.
Match confidence belongs beside the identity or relationship it describes. It should not be presented as overall decision confidence. Similarly, an opportunity score should be shown only when the evidence that feeds it meets the relevant eligibility rules. A large number of contextual observations should not compensate for a missing subject anchor.
4. Preserve provenance, dates, geography, and units
Every material value should be traceable to a source observation or an explicit assumption. The evidence envelope should retain the source family and row reference, subject, scope, effective date, observed date, release, geography, period, value, unit, match method, freshness and revision.
These attributes change interpretation. A sold-price median for a postcode district is not the selected property's sale price. A rateable value may relate to a hereditament rather than the whole building. An EPC may cover one unit. A local-authority demographic trend is useful for place analysis but not a property-level demand measurement.
5. Treat source failure, absence, and inapplicability differently
“No data” often hides several operational states. A source may not have been queried, the request may have failed, the returned records may not match, the evidence may be stale, the field may not apply, licensing may prevent display, or a suitable value may genuinely be absent. Each state implies a different next action.
| Display state | Meaning | Appropriate next action |
|---|---|---|
| Not attempted | The route has not run | Run it when the question justifies the cost and scope |
| Source unavailable | The route failed or timed out | Keep existing evidence visible and retry without claiming absence |
| No exact match | Sources were checked but no eligible property row matched | Show candidates and the matching boundary; do not promote a nearby record |
| Candidate review | One or more plausible records need analyst confirmation | Compare identifiers, addresses, dates and conflicts |
| Stale | An observation exists but falls outside the freshness policy | Display the dated value and request an update where material |
| Not applicable | The field does not apply to this asset or decision | Exclude it from the required-evidence denominator |
| Withheld | A value exists but policy or licensing prevents display | Explain the restriction without implying that the value is missing |
| Unresolved | No eligible answer is linked to a decision-relevant field | Create a specific verification gate with an owner and consequence |
6. Reconcile conflicting evidence before creating findings
Reconciliation should occur before the Decision Brief, scenario analysis and report are composed. Exact property identifiers should take priority over geographic context; effective and observation dates should guide recency; authoritative warehouse sources should take priority over approved enrichment or derived projections for the same direct fact; and competing values should be preserved rather than selected by source order.
Reliance limits can then reference stable gap identities. When every required field has eligible evidence, the limit can close automatically. The same gap should not reappear as repeated prose under several findings.
7. Apply the boundary consistently across screens, scenarios, and reports
A property fact should have one detailed home. Other views can refer to it without recalculating or restating it. The property workspace, decision evidence, property analysis and report should all reference the same immutable snapshot so that the conclusion does not change when the user moves between views.
Scenario assumptions must remain classified as observed, benchmark, derived or analyst-entered. Contextual benchmarks can support low, base and high cases when their geography, period, sample and derivation are disclosed, but they do not become property facts. A generated report should reproduce the same evidence scope and limitations as the screen.
8. Use plain language without weakening governance
Technical classifications should be translated, not hidden. “Candidate review” can be explained as “a plausible match was found, but the exact property or party still needs confirmation.” “Contextual” can be explained as “this describes the surrounding market or location rather than the property itself.”
Changing “unresolved” to a more reassuring label does not improve the analysis. A gap closes only when the required field is supported by an eligible source linked to the correct subject. The useful response is to explain why the boundary exists, what information remains available, and which action would resolve it.
A decision-ready evidence checklist
- Confirm the selected property using an eligible identifier or reviewed source-backed match.
- Attribute owner, occupier, operator and related-company evidence to the correct role.
- Show the scope, date, geography, period and unit for every material statistic.
- Separate confidence, coverage, match confidence and opportunity scoring.
- Keep candidate, contextual, stale and conflicting evidence visible but outside ineligible conclusions.
- Distinguish source failure from a genuine absence of matching data.
- Record explicit assumptions and do not present them as observations.
- Use the same immutable evidence revision across analysis and reporting.
- Turn material gaps into named verification actions with a decision consequence.
A well-bounded review lets another person trace the conclusion to the evidence, understand the counter-case and see what must be checked next. That makes the analysis easier to challenge, reproduce and improve.