Measurement
Part of Measuring property media with metric contracts, governed tracking and honest attribution
Four ways to attribute a property enquiry, and how much each one can prove
Compare four England property-media attribution methods by event coverage, identity needs, assumptions, bias, privacy, review and causal strength.
Attribution methods for property creators answer which content receives credit under a chosen rule. They do not, by themselves, show that the content caused an enquiry, instruction or property outcome. No method in this comparison proves causation. For England property media, choose the least intrusive method that answers the real decision.
Comparison method: on 5 September 2026, the editorial research team compared four method families on the same criteria: counted event, population coverage, identity requirement, assumptions, likely bias, privacy burden, reproducibility and causal strength. No vendor, channel account or dataset was tested. Costs and product features are excluded. The decision rule below maps decision type to a recommended method.
What to take away
- Attribution methods assign credit under a chosen rule but do not prove that content caused an enquiry.
- Tagged links miss copied links, cross-device activity, offline contact and deleted identifiers, so can over-credit.
- Self-reported source is limited by recall, interpretation and social-desirability bias, and may not identify a specific asset.
- Enquiry-system campaign fields require consistent population and privacy review.
- Controlled impact evaluation estimates difference from a counterfactual, but small samples and selection can make it inconclusive.
Tagged link or content code
A tagged link assigns an eligible observed destination event to a content ID within a declared period. Define the event, property and content population, lookback window, timezone, duplicate rule, internal-test exclusion, source system and query version. The owner reports matched and unmatched events.
Tagged Link Attribution Steps
- Define event, property and content population
- Set lookback window and timezone
- Apply duplicate rule and exclude internal tests
- Record source system and query version
- Report matched and unmatched events
This method assumes the tag survives and the same rule fits the journey. It misses copied links, cross-device activity, offline contact and deleted identifiers. It can over-credit the selected touch.
For example, a tagged link in an Instagram reel for a Leeds flat records a click to a booking page, but a link copied into a WhatsApp group is not counted. If storage or access technologies are involved, the ICO's 2026 guidance requires assessment of the actual technology and purpose.
Self-reported source
Ask an adult enquirer how they first heard about the service using a neutral, optional field. The event is a completed answer; the population is eligible completed enquiries during the fixed England period. Report each response count over all eligible enquiries, including "not answered" as a visible category. Record question version, order and coding owner.
For example, an enquirer for a Bristol terrace says they first heard about the agency on Instagram, which may not identify the reel.
Recall, interpretation and social-desirability bias limit the result. A response such as "social media" may not identify a particular asset. Do not join free text to personal profiles merely to improve matching. ICO data-minimisation guidance requires personal data to be limited to what the purpose needs.
Enquiry-system campaign field
A controlled campaign code carried into the enquiry record can connect an observed submission to an approved release. Define eligible forms, accepted codes, join key, clock, duplicate handling, missing-code treatment and query release. The denominator is all eligible submissions, not only successfully matched ones.
For example, a flyer code from a York open house is typed into a portal enquiry form and lands in the customer record.
This route assumes the code is populated consistently and the enquiry system records the intended event. Staff edits and imports can change it. Access, retention and processor arrangements need review under UK GDPR, PECR and ICO guidance. An attributed enquiry is still not a transaction, a property value or incremental contribution.
Controlled impact evaluation
An experiment or suitable quasi-experimental comparison estimates the difference between an intervention population and a credible counterfactual. Pre-specify allocation, outcome, denominator, period, contamination, exclusions, sample requirements, analysis, uncertainty and stopping rules. A qualified evaluator owns the design and any causal language.
For example, compare enquiry rates in two matched postcode areas in Birmingham, one exposed to a campaign and one not.
The 2026 Magenta Book explains that experimental and quasi-experimental approaches use comparison data to infer impact, subject to design assumptions. Its quality guidance says monitoring change alone does not establish causation. Small samples, selection, concurrent market changes and spillover can make a design inconclusive.
Choose and report
Use a crediting method for operational allocation and a causal design only when the decision warrants it. Publish the exact event, population, numerator, denominator, unit, Europe/London period, sources, query version, inclusions, exclusions, owner and uncertainty. Keep platform observations, attributed outputs, causal estimates and editorial interpretation in separate columns.
| Decision type | Recommended method |
|---|---|
| Allocate content budget or credit a channel | Tagged link or content code |
| Understand first reported touch | Self-reported source |
| Match a campaign release to an enquiry record | Enquiry-system campaign field |
| Estimate incremental impact | Controlled impact evaluation |
Advertising, FCA, privacy and PECR, security, accessibility, rights and property reviews remain independent. Better attribution cannot cure a hidden ad, unlawful collection, unsupported property statement or unapproved financial promotion.
Before you act
- Define the counted event, population and lookback window.
- Report matched and unmatched events with denominators.
- Include not answered as a visible category.
- Pre-specify allocation, outcome and stopping rules for experiments.
- Publish event, population, numerator, denominator and uncertainty.
- Keep platform observations and causal estimates in separate columns.
Common questions
What does a tagged link method actually measure?
A tagged link assigns an eligible observed destination event to a content ID within a declared period. It assumes the tag survives and the same rule fits the journey. It misses copied links, cross-device activity, offline contact and deleted identifiers, and can over-credit the selected touch.
Why is self-reported source not a reliable attribution method?
Recall, interpretation and social-desirability bias limit the result. A response such as social media may not identify a particular asset. The method requires a neutral, optional field and reporting each response count over all eligible enquiries, including not answered as a visible category.
When does an enquiry-system campaign field prove causation?
It does not. A controlled campaign code connects an observed submission to an approved release. It does not show that the campaign caused the enquiry.
