Measurement
Part of UGC marketplaces measurement: keeping six evidence classes apart
UGC marketplaces measurement mistakes: six UK-wide checks
UGC marketplaces measurement mistakes: six UK-wide checks for attribution, metrics, corrections, privacy, ad claims and rights on work bought in England.
What to take away
- The six UGC marketplaces measurement mistakes covered here are a moving eligible population, platform credit read as incremental, incompatible units added together, overwritten corrections, reused personal data, and a chart that overrides a failed gate.
- Each defect maps to a named UK-wide controlONS quality guidance, HM Treasury's QPIE supplement, the OSR Code principles, ICO design guidance, CAP Code section 2 and the Copyright, Designs and Patents Act 1988.
- Avoidance is a reporting convention, not a new analytics stack. Fix the definition, keep the prior version, and label platform credit as attribution.
- Reach, spend or a higher engagement figure cures none of the six.
Scope, cut-off and inclusion rule
Research closed on 6 September 2026, and scope covered official UK statistical, evaluation, privacy, advertising and rights sources. These UK-wide sources are applied as methods or controls to one UGC asset bought in England; they are not England-specific requirements.
If a buyer has sourced creator work through Cohley or IZEA, these checks apply to that campaign record: compare like-for-like deliverables, keep platform-attributed views separate from buyer outcomes, and track usage permission for each asset. This is an application example, not a claim about either service's own dashboard or a comparison of providers.
A mistake entered the list when a named source supported the exact evidence boundary, and when the defect could change a commission decision. Supplier benchmarks, invented results and legal conclusions were excluded. The sequence is not a ranking.
Two dated anchors run through the list: the 6 September 2026 cut-off, and the Copyright, Designs and Patents Act 1988, still the primary UK statute for copyright and usage rights in creative works.
The six mistakes and the control for each
| Mistake | How to avoid it, and the named control |
|---|---|
| 1. The eligible population changes between periods | Fix the event, identity, period and exclusions before you query. Control: ONS, Quality in official statistics: quality defined. |
| 2. Platform credit reported as incremental effect | Use attribution wording unless a counterfactual design exists. Control: HM Treasury Magenta Book, QPIE guidance. |
| 3. Incompatible observation units added together | Keep the dated source definition and show not comparable. Control: OSR, Putting the Code principles into practice. |
| 4. A correction overwrites the earlier total | Retain the prior version and log the reason for revision. Control: OSR, Putting the Code principles into practice. |
| 5. Personal data reused because it is available | Record purpose, fields, access, joins, retention and deletion. Control: ICO, Data protection by design and by default. |
| 6. A positive chart overrides a failed gate | Keep recognition, substantiation and permissions separate. Control: CAP Code section 2; Copyright, Designs and Patents Act 1988. |
A numerator for accepted releases cannot use all drafts as the denominator one month and published assets the next. ONS describes coherence and comparability inside its fitness-for-purpose quality definition. That page is a voluntary method reference, not a claim to official-statistics status.
A platform label records its own attribution method and its own visibility. It cannot show what would have happened without the asset. HM Treasury's QPIE guidance separates monitoring change from causal impact through a counterfactual. Our note on attribution methods explains why one asset rarely settles that question.
Views, impressions, interactions, people and buyer outcomes are different units, and matching labels can still hide different windows or duplicate rules. The OSR's guidance on Putting the Code principles calls for rigour and openness about quality. Mark the result not comparable when the fields differ.
Overwriting a dashboard total hides whether a withdrawal, return, duplicate or source change moved the number. The principle worth borrowing is traceability: keep the earlier version and log why it changed.
A campaign export should not become an audience profile or an outreach list by convenience. The ICO's data protection by design and by default guidance covers purpose limitation and minimisation across the lifecycle. A qualified reviewer decides the lawful basis.
Reach and attributed activity cannot cure unsupported product wording, hidden commercial content or missing rights. CAP Code section 2 keeps recognition and substantiation as distinct requirements.
The Copyright, Designs and Patents Act 1988 sets the permission boundary for the commissioned asset; an IPO ownership record is only a starting point. Checking advertising rules tests recognition, claims and rights on a single advert.
Example: one correction that flatters a total
Illustrative figures follow. A dashboard shows 40 accepted creator assets on day 1 and 44 on day 12, and the earlier cell was overwritten in place.
Nobody can now tell whether four assets arrived or a duplicate rule changed. Keep a dated copy, log the revision reason, and publish both readings with the date each was taken.
Before you act
- Define the event, identity, period and exclusions before you query.
- Use attribution wording unless a counterfactual design supports causal language.
- Keep the dated source definition and mark results not comparable when fields differ.
- Retain the prior version when you correct a dashboard total.
- Record purpose, necessary fields, access, joins, retention and deletion.
- Treat recognition, substantiation and permissions as separate gates.
Common questions
These are provider-neutral controls: apply them to the campaign record whether a creator asset was sourced through a marketplace or commissioned directly. They do not establish that every marketplace defines a metric in the same way.
Why is a platform label not evidence of incremental effect?
It records the platform's own attribution method and visibility, not what would have happened without the asset. QPIE separates monitoring change from causal impact through a counterfactual, so report rule credit as attribution.
What happens when observation fields differ?
Views, impressions, interactions, people and buyer outcomes are different units. Preserve the dated source definition and show not comparable instead of adding the rows together.
Is any of this regulated for a private UGC report?
No. The ONS, OSR, ICO and CAP documents are methods and codes you can adopt voluntarily. A qualified reviewer decides any legal question about personal data or rights.
