How to Measure a PR Campaign: A Worked Example
- How do you measure a PR campaign?
- What belongs in the measurement plan?
- What is the difference between outputs, outtakes and outcomes?
- How can you assess coverage without relying on volume alone?
- What does a worked report look like?
- How do tracking links help, and where do they stop?
- How do you avoid overstating causation?
- Should you report advertising value equivalents?
- What privacy checks belong in the workflow?
- How should the final report end?
- Sources
How do you measure a PR campaign?
Measure a PR campaign against a defined audience objective, using a documented baseline, reporting period and evidence source. Separate coverage and other outputs from audience responses, outcomes and longer-term impact. Report what the data supports, including missing evidence and competing explanations. Do not turn media mentions into a revenue claim or treat a tracking link as proof that PR caused a sale.
A useful report answers a decision: what should the team continue, change or investigate? The method below is an editorial reporting workflow for a small communications team, not a claim that every campaign needs the same dashboard.
AMEC's Barcelona Principles 4.0 call for measurable objectives, stakeholder understanding, relevant channel coverage, qualitative and quantitative analysis, and transparent methods. Start there, rather than with whichever export has the largest number.
What belongs in the measurement plan?
Write the reporting question before choosing a metric. For example: did the announcement help eligible organisations understand the new service's availability, or did it simply produce coverage?
Create a short measurement note alongside the PR brief. Our suggested fields are:
- Audience: the specific people whose understanding or behaviour matters.
- Intended change: the question or action the communication should address.
- Evidence: the data source that can actually observe that change.
- Baseline: the comparable starting observation, or an explicit statement that none exists.
- Window: when measurement begins and ends.
- Owner: who maintains the data and explains its limitations.
- Decision: what a useful or disappointing result would change.
These are working fields, not a substitute for research design. If awareness is the objective but the only available data is website visits, record the gap. Do not relabel visits as awareness.
Keep planning targets distinct from observed results. A target agreed after the campaign finishes is not a pre-campaign success criterion.
What is the difference between outputs, outtakes and outcomes?
AMEC's evaluation taxonomy separates communication distribution, audience response, audience effects and wider results. Outtakes can overlap short-term outcomes, and outcomes can overlap impact.
| Level | What it asks | Example evidence |
|---|---|---|
| Output | What communication reached the intended setting? | Relevant coverage and message inclusion |
| Outtake | What did people notice or understand? | Recall or understanding responses |
| Outcome | What changed in the audience? | Trust, intention or qualified registrations |
| Impact | What wider result did communication contribute to? | Organisational or stakeholder results supported by additional evidence |
These examples are illustrative. Stated intention is not completed action. Keep your counting definition stable; flag changes rather than treating differently defined counts as comparable.
How can you assess coverage without relying on volume alone?
Use a coverage register that lets another reviewer reproduce the count. Our proposed record contains the URL, publication date, outlet, relevant audience, original or syndicated status, material messages included, factual errors, and review notes.
Define inclusion rules before tallying. Does a passing mention count? Are duplicate URLs counted once? Will syndicated copies be reported separately from original stories? There is no useful denominator until those questions have answers.
For message inclusion, record what the article actually says. A link to the announcement is not evidence that a specific message appears in the story. If an item inaccurately describes eligibility, record the error even if the surrounding tone is positive.
Keep qualitative notes alongside the count. A relevant explanation, a repeated misunderstanding and a correction request require different responses. A single overall sentiment label may not tell the team which response is needed.
For social versions of the same announcement, preserve a shared campaign label while recording each asset's specific job. Do not pool an explanatory post and a registration reminder merely because both use the same image.
What does a worked report look like?
Consider this entirely fictional service-announcement campaign. None of the figures below is an industry benchmark or a real client result.
The team defined 24 distinct, relevant coverage items after its documented exclusions. Eighteen included the essential availability message. Its report therefore shows:
- Message inclusion: 18 / 24 = 75%.
- Items without that message: 24 - 18 = 6.
A separate enquiry record contains 40 submissions. Eight are repeat submissions, leaving 32 unique enquiries. Of those, 20 meet the team's predefined eligibility criteria:
- Unique enquiries: 40 - 8 = 32.
- Eligible share of unique enquiries: 20 / 32 = 62.5%.
The other 12 unique enquiries remain in the report as not meeting those criteria. They should not disappear because they are inconvenient.
These are different denominators. The 75% describes reviewed coverage; the 62.5% describes unique enquiries. Combining them into an average "campaign score" would have no defined interpretation in this example.
Suppose a comparable prior period had 16 eligible enquiries. The change to 20 is four enquiries, or 4 / 16 = 25%. That is an observed increase. It is not evidence that the campaign generated all four additional enquiries.
How do tracking links help, and where do they stop?
In Google Analytics, campaign parameters can identify the campaign associated with referral traffic. Google's URL-builder documentation describes fields including utm_source, utm_medium and utm_campaign; utm_content can distinguish creative variants.
Agree a naming convention and use it consistently. For example, an editorial naming record might use campaign service-briefing and content labels availability-explainer and registration-reminder. These are invented labels, not a recommended analytics configuration.
Have the analytics owner test the implementation and interpret the relevant report. Google distinguishes user-, session- and event-scoped traffic-source dimensions. First-user acquisition, session acquisition and credit assigned to a key event answer different questions. State which one the report uses rather than calling every figure "PR conversions."
A tracking label helps identify a recorded path. It does not identify every influence on the person's decision, or establish what that person would have done without the campaign. Missing source information is a limitation to report, not permission to assign the result to PR.
How do you avoid overstating causation?
Ask what else changed: pricing, sales outreach, availability, paid promotion, seasonality or another announcement. Keep a dated context note alongside the report.
The UK government's quality-in-impact-evaluation guidance distinguishes observing a change from establishing responsibility for it. It explains why a simple before-and-after comparison without a credible counterfactual usually provides weak causal evidence. A counterfactual is an estimate of what would have happened without the intervention.
Applied to the fictional enquiry example, the defensible statement is narrow: "Eligible enquiries increased from 16 to 20 across the compared periods; the available records do not isolate the campaign's effect."
If a major spending decision depends on causal impact, involve a qualified evaluation specialist before launch. A later spreadsheet cannot create a comparison or recover evidence that was never collected. Avoid claiming that an informal comparison amounts to a controlled study.
Should you report advertising value equivalents?
No. AMEC's Barcelona Principles 4.0 reject advertising value equivalents, commonly abbreviated AVEs, as invalid measures. They call instead for evaluating communication through outcomes and impact.
Do not replace an AVE with a different unsupported currency figure. An estimated audience, a mention count or a hypothetical advertising rate is not revenue attributable to the campaign.
If the evidence supports only coverage analysis and enquiry counts, report those accurately. A limited report with transparent definitions is more useful than a financial claim that cannot be reconstructed.
What privacy checks belong in the workflow?
Collect only information the evaluation genuinely needs and have qualified privacy or legal staff review the collection method, permissions, access and retention arrangements. This guide is not legal advice.
Google's Analytics privacy guidance prohibits sending information Google could recognise as personally identifiable, including email addresses and personal mobile numbers. It specifically warns against putting such information in campaign parameters.
Use campaign-level labels, not a journalist's name, a customer's email address or sensitive enquiry details. Keep any authorised identifiable operational record access-controlled and separate from a broadly circulated aggregate report. Do not upload private interview notes or customer records to an unapproved analysis tool.
How should the final report end?
Close with a decision and its evidence, not a list of exports. Our suggested final block has four lines: finding, limitation, next action and owner.
In the fictional example, a reasonable next action is to examine why six coverage items omitted the availability message and why 12 unique enquiries fell outside the defined criteria. That does not assume every omission has the same cause.
Keep the underlying inclusion rules and calculations with the report. If a factual error is discovered, correct the affected finding and record the change. The purpose of measurement is to improve decisions without making claims stronger than the evidence.