PR agencies measurement: reject AVE, keep evidence levels apart

PR agencies measurement begins with a decision, not a coverage total. An organisation operating in the UK should decide what it needs to learn, who can act on the finding and what evidence would change the next step. Only then should it select measures, data sources and a reporting rhythm.

This guide sets out a system rather than a set of results. It contains no campaign findings and no UK industry benchmark. Many PR measurement mistakes start with an undefined measure, so every source, definition and collection route should be checked against the buyer's own data before a value enters a report.

What to take away

Start with the decision record

A useful brief names the organisational decision, eligible audience, starting condition, intervention, delivery period, evidence owner and review date. It also records the current or no-activity alternative. Without that baseline, a later movement may be visible but its cause is harder to assess.

Write one question per decision. For example, a delivery owner may ask whether approved material reached the selected media-contact population. A communications lead may ask whether a defined stakeholder group noticed and understood it. A finance owner may ask whether an observed organisational change justifies further spending. These questions sit at different levels and cannot share one convenient answer.

The AMEC evaluation taxonomy is global professional guidance rather than UK law. It separates inputs, activities, outputs, audience outtakes, outcomes and organisational or stakeholder impact. Its examples are prompts, not default metrics or UK norms. Use the distinctions to stop a media item from being reported as an audience or business effect.

Agree the system with your PR agency

The measurement system belongs to the client, whoever is appointed. During selection, ask each shortlisted agency to describe how it would evidence each level and who would own the definitions, the source data and the correction history.

Require a written metric contract for every reported value, including the source system, query version, missingness and a named owner on both sides. An agency should say what it cannot evidence as clearly as what it can. No report should present advertising value equivalent, or a renamed variant, as earned-media value, impact or return on investment.

Keep the definitions, data routes and correction history under the client's control, and agree an exit route for data when the contract ends. A directory listing, a credentials deck or a supplier case study is an attributed claim, not evidence of effect.

Keep the evidence levels apart

Inputs and activities

Inputs are resources and prior evidence. Activities are the work performed, such as research, drafting, approval, media relations or event delivery. A budget line, hours record or completed briefing shows use of resource. It says nothing by itself about reception or effect.

The record should identify the unit, period, currency where relevant, VAT treatment, source, owner and uncertainty. Contracted activity remains contracted even if it was never completed. Observed activity requires an operational record. Do not merge the two.

Outputs

An output is a delivered communication or documented exposure in a defined population. Count the exact artefact, state if the channel was owned, paid, shared or independent editorial.

The agency can evidence a release sent, briefing held or approved item published on a controlled channel. It cannot make an independent editor publish, select a headline or keep article online.

Coverage counts need inclusion rules for duplicates, syndication, updates, inaccessible pages and content outside the chosen geography or period. Potential audience figures are not observed attention. A publisher or monitoring supplier's label remains an attributed source claim until independently checked.

Audience response and outtakes

Attention, recall, understanding, consideration and other immediate responses require evidence from the people or behaviour in scope. Page delivery, impressions or a clipping do not show what anyone understood. A survey needs a population, sampling frame, recruitment method, field dates, question wording, response pattern, weighting, missingness and uncertainty. Behavioural traces need their own collection and interpretation limits.

Outcomes

An outcome is a defined change in the audience, such as corrected understanding or a completed action. It needs an eligible population, baseline, period and rule for counting each person or event. The organisation must decide whether the outcome was merely observed, allocated by an attribution rule, modelled or estimated causally.

Organisational impact

Impact belongs to the organisation or stakeholders, not a media clipping. It could concern a finance-owned contribution measure, service demand, staff retention, public behaviour or another stated objective.

The Barcelona Principles 4.0, published by AMEC for global communications practice, calls for reporting outputs, outcomes and impact while rejecting advertising value equivalents. It does not demonstrate that a specific PR activity caused any result.

Give every metric a contract

Before a value enters a dashboard, complete these fields:

The definition travels with the value. A percentage without its denominator is incomplete. An elapsed time without a clock is not comparable. A sentiment label without coding rules, language coverage, validation and an uncertainty note is only a vendor or analyst classification.

Record the value's evidence class

Keep separate fields for values that are:

Never overwrite one class with another. A platform may assign an enquiry to a channel, but that allocation does not establish incrementality. A finance forecast is not realised value. An outcome change is not a causal effect unless the design supports that conclusion.

HM Treasury's updated Magenta Book is written for government evaluation. Its split of process, impact and value-for-money questions is useful beyond government, if carefully adapted.

Its companion quality guide for impact evaluation notes that monitoring can show a change without showing the intervention caused it. A qualified analyst should select and review any causal design used for a commercial PR decision.

Reject AVE and its lookalikes

Advertising value equivalent applies an advertising space or time price to earned coverage, sometimes with a multiplier. That arithmetic does not become earned-media value, impact or return on investment. Barcelona Principles 4.0 identifies AVE and similarly renamed variants as invalid measures.

Do not rescue the method by changing its label. If a buyer has a separate operational reason to estimate media space cost, name the estimate, formula and limitations without using the words value, impact or ROI. Keep it outside outcome and financial contribution fields. Independent editorial content has different control, context and audience response from purchased advertising.

Return on investment needs a finance-approved numerator and full economic cost denominator for the same scope and period. Attribution and causation remain separate. The calculation should stop when revenue contribution, cost, counterfactual or uncertainty cannot be defended.

Build a dashboard that shows data health

A dashboard is a controlled view of evidence, not the underlying evidence. Its header should state the decision, population, geography, period, timezone, refresh time, source versions and owner. Each card should expose the metric definition and evidence class. Keep targets blank until the buyer approves a baseline and tolerance.

Show freshness, missingness, duplicate handling, suppressed values, revisions and open disputes beside results. The ICO's data-minimisation guidance says personal data should be adequate, relevant and limited to what is necessary for the purpose. A dashboard therefore should not become a convenient store of contact-level detail.

Tracking technologies require a separate permission and transparency assessment. The ICO's final storage and access technologies guidance was updated on 29 April 2026 and covers technologies such as cookies, pixels and device fingerprinting under PECR and, where relevant, UK GDPR. It does not approve a particular analytics configuration or prove data completeness.

Access and revision evidence matter too. The NCSC's cloud audit-information guidance is security guidance for choosing and using cloud services. It supports checking the availability, format and retention of logs, but it is not a measurement-quality certification for a PR dashboard.

Make the report usable without hiding limits

Give each decision owner the smallest view that answers their question. The delivery team may need item-level workflow exceptions. A director may need trend, uncertainty and the decision due. A board may need the theory, outcome evidence, financial boundary and residual risk. Do not expose personal data simply because the reporting system can.

Accessible reporting is part of the work. The government's inclusive communication guidance was written for government communicators. Its prompts on formats and channels can inform a buyer's design, but a named accessibility reviewer must decide and test the actual private-sector report.

Every chart needs a text equivalent, defined reading order and legible labels. Colour should not carry meaning alone. Tables should retain units and source notes when exported. A correction log must identify the original value, replacement, reason, authorisation and every report affected.

Handle benchmarks as a research question

A benchmark is comparable only when population, geography, service definition, channel, period, unit, numerator, denominator, collection method and treatment of missing data match closely enough for the buyer's decision. Figures for the UK as a whole, for individual nations and regions, and for global markets stay in separate rows. Supplier case studies remain attributed claims. A media average from a different audience or campaign type cannot fill a gap in UK evidence.

If compatible external evidence does not exist, publish that gap. Use the organisation's own version-controlled baseline, or commission a properly designed study. Do not pool reach, engagement, share of voice, sentiment, conversion or ROI figures merely because they share a label.

Govern correction and retirement

Name an owner for collection, calculation, interpretation, privacy, security, accessibility, rights and finance. These are separate gates. The claims reviewer checks what the report says; the analyst checks the method; the business owner decides what to do. No composite score can compensate for a failed legal or evidence control.

Pause a measure when its source changes silently, the denominator cannot be reconstructed, missingness exceeds the buyer's blank tolerance, permissions lapse, the query fails reconciliation or the definition no longer matches the decision. Reissue affected reports after correction. Retire vanity measures that prompt no decision or repeatedly invite a claim they cannot support.

Recheck the AMEC, HM Treasury, ICO, NCSC and accessibility sources, plus every platform method and contract used in the actual system, whenever a source or a platform method changes. Before the system goes live, named qualified reviewers should inspect the buyer's definitions, data routes, calculations, permissions, tests and correction history.

Before you act

Common questions

Why does the guide reject advertising value equivalent?

AVE applies an advertising price to earned coverage, sometimes with a multiplier, but that arithmetic does not become earned-media value, impact or return on investment. Barcelona Principles 4.0 identifies AVE and similarly renamed variants as invalid measures. Renaming the method does not rescue it, and any separate media space cost estimate should stay outside outcome and financial contribution fields.

What must a metric definition include before it enters a dashboard?

Each metric needs a purpose, event or variable, eligible population, numerator and denominator, unit, period and timezone. It also needs inclusions and exclusions, source system and query version, missingness, ownership, uncertainty, correction rules and a retirement trigger. The definition travels with the value, so a percentage without its denominator is incomplete.

How should outputs be counted and reported?

An output is a delivered communication or documented exposure within a defined population. State whether the channel was owned, paid, shared or independent editorial, and apply inclusion rules for duplicates, syndication, updates, inaccessible pages and content outside the chosen geography or period. Potential audience figures are not observed attention.