Reporting on a creator media kit without inventing impact

Creator media kits measurement starts with a record of what was sent and ends only where the evidence supports an interpretation. Keep distribution, recorded interaction, commissioning enquiries and agreed work separate. Decide the event definitions and eligible population before calculating a proportion, and retain the limitations when a figure moves into the portfolio itself.

This 2027 planning guide proposes a reporting design for creator media kits and portfolio tools used in England. Its evidence review is dated 6 September 2026; it contains no future observations, platform tests or measured creator results. The design must be adapted to the actual records and reviewed by a fact-checker and measurement reviewer before use.

What to take away

Contents

Start with the decision about the kit

The deliverable is a four-register record plus a bounded statement.

Write down what the owner needs to decide. A question about whether the current portfolio reached its intended recipients needs distribution evidence. A question about whether a revised work-sample page changed commissioning behaviour needs an evaluation design. They may use some of the same records, but the conclusions are different.

Give the question a version identifier. Record the content revision, the intended audience and the point at which that version became the candidate for distribution. If the design keeps changing during observation, preserve the changes and decide whether the results still describe one intervention.

The Magenta Book distinguishes monitoring from evaluation and sets out the role of evidence in assessing impact. It is central government guidance, not a certification scheme for private portfolios. Its distinction supports the house rule here: routine activity records should not be presented as proof of additional business caused by a kit.

Avoid selecting the reporting question after seeing which number looks strongest. Keep an agreed question in the record even if the first result is unavailable or unhelpful. An unresolved measurement problem may be the useful finding for the owner.

Define who belongs in the report

Choose the unit that fits the question. For a commissioning enquiry study, the proposed unit is a recipient organisation, because several contacts may represent the same potential commissioner. This is a design choice. Record how contacts are grouped and what happens when the organisation cannot be established.

State how England coverage is determined. One possible house specification uses the organisation's recorded commissioning location. Do not infer that location from the creator's address, an email suffix or a UK-wide label. Keep unknown locations separate and describe any mixed-geography records.

Specify the cohort entry rule before distribution. Identify the eligible recipients, the relevant kit version and whether existing clients are included. If existing negotiations are outside the question, document that exclusion before examining outcomes. Preserve a separate record of excluded cases and the reasons.

Choose an observation period with start and end dates, a timezone and a follow-up endpoint. There is no universal duration supplied by this guide. The owner and analyst must justify a window that fits the intended question and available evidence, then apply it consistently.

Build a dictionary that another person can use

For every measure, write the event trigger in ordinary language. Include its unit, source, source owner, numerator and denominator where relevant, applicable dates, cohort, exclusions and known uncertainty. An event name without a trigger does not tell a reviewer what was counted.

ONS's quality definitions include relevance and coherence or comparability. For the proposed portfolio dictionary, use those concerns to ask whether a field fits the decision and whether its meaning survives a comparison. The source supplies quality concepts; the record structure below is an editorial proposal.

Worked example, illustrative only: event, approved version released; unit, recipient organisation; source, distribution register; owner, release coordinator; numerator, 42 organisations; denominator, 60 scheduled organisations; missing-data label, coverage incomplete for 5 scheduled organisations. The counts show the record structure and are not a benchmark.

Assign owners to the records they can actually inspect. Do not put a creator's name against an export controlled by somebody else without resolving access and responsibility. These proposed responsibilities make the evidence traceable; they do not prescribe a team size.

Give distribution its own measure

Define a distribution event as the recorded release of the approved version to an eligible organisation. Retain the destination category and release time in the distribution register. Have the release owner distinguish an initial release from a corrected copy or replacement.

A house distribution-completion proportion would use organisations with that recorded release as the numerator and organisations scheduled to receive that version as the denominator. Both must belong to the same declared cohort and distribution window. Record cancelled destinations under the agreed exclusion rule.

This measure describes the completeness of the release process. It does not establish that a message arrived, that a document opened or that someone considered the work. If the register is incomplete, record the coverage issue instead of assuming the missing rows were successful releases.

Use this administrative view to locate missing records or an unfinished handover. Avoid interpreting a change in the proportion as improved persuasion. That would require a different event and supporting evidence.

Record interaction only where it is observable

For a hosted portfolio, the operator must identify the actual event specification before reporting engagement. Record the event trigger, export source and any matching mechanism. For an exported file, do not assume that reading or forwarding can be observed. No platform behaviour has been tested for this guide.

If an interaction can be linked reliably to an eligible organisation, define a proportion using organisations with at least one qualifying event over organisations with a documented opportunity to interact in the same window. Keep event counts separate from organisation counts. Repeated activity is not automatically another potential commissioner.

Document test activity, known duplicates and other exclusions that the actual evidence can support. If automated and human activity cannot be distinguished, describe that limitation. Do not claim a cleaning procedure removed a category unless it was implemented and checked.

Place unmatched activity outside the recipient proportion. It may remain useful as a separate observed count, provided its definition and limitations are visible. A partial matching process cannot justify a claim about everyone who saw the portfolio.

Classify enquiries and agreements independently

Adopt a written commissioning-enquiry definition, such as a received message requesting discussion of a potential piece of work. Let the enquiry-register owner apply that rule and record ambiguous cases. Keep spam, supplier pitches and existing project administration outside the proposed measure, with reasons.

Count an organisation's first qualifying enquiry in the follow-up period for a recipient-level response proportion. Use all eligible organisations in that distribution cohort as the denominator, subject to the declared study rules. Do not restrict the denominator to organisations known to have replied.

An agreement needs a separate documented event in an agreement register. Specify what the authorised owner must see before marking work agreed. This guide sets no accounting recognition rule or revenue valuation.

If an agreement proportion is needed, apply the same eligible-cohort discipline and a declared follow-up endpoint. Track pre-existing negotiations and withdrawn agreements under explicit rules. A commission after distribution can be recorded without asserting that the kit generated it.

Reconcile the sources before presenting a percentage

Ask the analyst to trace a proposed numerator back to the supporting records. Check that records belong to the chosen version, population and period. Retain the extraction date and any revisions, so another reviewer can reconstruct the calculation from the same inputs.

Write a proportion as qualifying eligible units divided by all eligible units in the specified population and window, multiplied by 100. Show the two counts beside the percentage. This is a calculation definition, not an illustrative result or recommended conversion rate.

Worked example, illustrative only: 42 organisations with a recorded release divided by 60 scheduled organisations gives 70 per cent, labelled coverage incomplete for 5 scheduled organisations. If a scheduled organisation has no recorded release status, report it as missing rather than as a non-release.

If the denominator is zero, the proportion is undefined. If the source is unavailable, the observation is missing. Those states need different labels from an observed zero numerator.

When comparing periods, check event definitions and follow-up lengths before calculating change. Preserve a break note if the collection method changes. Do not force incompatible observations into a continuous series merely to make the chart look complete.

Describe uncertainty without inventing precision

ONS explains sampling and non-sampling errors, including inaccurate responses and processing problems. The relevance to portfolio reporting is that a complete-looking extract may still have classification or coverage limitations. The source does not provide a ready-made statistical interval for a creator's administrative records.

List the uncertainties that actually apply: unknown recipient identity, incomplete follow-up, disputed enquiry classification or an unverified collection change. Ask a qualified analyst whether any numerical uncertainty estimate is appropriate. A confidence interval should have a defensible method and interpretation, not merely software output.

ONS chart guidance recommends displaying uncertainty when it changes interpretation. If the evidence cannot sustain a meaningful comparison, explain the gap rather than presenting a precise-looking trend.

Keep source notes with exported tables as well as the dashboard. The receiving reader may never see the interactive screen or its hidden information panel. A number should remain interpretable outside the workspace where it was produced.

Decide whether attribution is feasible

A change in enquiry counts after a redesign is a descriptive observation. To investigate the redesign's effect, specify the intervention and an evaluation question before choosing a method. Consider recipient selection, other outreach, existing relationships and what the comparison would represent.

Magenta Book Annex A describes experimental, quasi-experimental and theory-based approaches. Its methods have different evidence requirements and limits. Choosing one for a kit requires a methodological assessment; a tracking tag or time sequence is not a substitute for that work.

Retain the administrative report even if attribution cannot be established. It may still show which records need repair or whether the intended distribution took place. Keep any causal language out of the portfolio claim until the evidence and interpretation receive qualified approval.

Admit benchmarks only after compatibility checks

No England creator-kit conversion benchmark exists. A future reference needs a source population, data owner, collection period, event definition and denominator. It must also explain exclusions, participation and uncertainty well enough to assess compatibility.

Distinguish an average across creators from a pooled rate across their recipients. Ask whether a selected customer sample represents the population named in the claim. Treat unknown geography or a different follow-up period as a reason to qualify or reject the comparison, not a detail to omit.

Keep the benchmark decision with its evidence. Record why a candidate source was excluded so it is not repeatedly reintroduced because its number is convenient. A methodological publication supplies a way to assess data; it is not itself a creator-performance dataset.

Release a report with a bounded statement

OSR's dashboard guidance emphasises user needs and maintenance for official statistical dashboards. Applied as a house principle, that means choosing a report someone can use and assigning responsibility for corrections. It does not import official-statistics release requirements into a private kit operation.

Before a figure becomes promotional copy, examine its wording separately. CAP's substantiation and qualification provisions address objective claims within the Code's remit. Keep dates, scope and material limitations with the statement, and obtain fact-checker and measurement reviewer sign-off where advertising interpretation is needed.

The proposed handover should contain the dictionary, retained extracts, calculation record, uncertainties and an approval decision. Identify the named author, fact-checker and measurement reviewer before publication. No such human sign-off is claimed here, and any actual personal-data design also needs appropriate privacy review. Recheck the sources at release and retain the article's hold until those decisions are complete.

Before you act

Common questions

What is the difference between a distribution measure and an interaction measure?

A distribution measure counts the recorded release of the approved version to eligible organisations. It describes the completeness of the release process. It does not show that a message arrived, a document opened or anyone considered the work. Interaction measures need an observable event and a reliable link to an eligible organisation.

How should the eligible population be defined for a commissioning enquiry study?

The proposed unit is a recipient organisation, because several contacts may represent the same potential commissioner. Record how contacts are grouped and what happens when the organisation cannot be established. State how England coverage is determined, keep unknown locations separate, and specify the cohort entry rule before distribution begins.

What must be recorded before an agreement is counted?

An agreement needs its own documented event in an agreement register, with the authorised owner specifying what they must see before marking work agreed.