Your coverage report survives the quarter and not the question. One sentence about how the number was built, and the page goes quiet. What would you put there instead, if the same coverage had to be priced against reach you could buy?
The slide has a screen full of placements on it and one number at the bottom. Someone asks what the number means. You explain that it estimates what the coverage would have cost as paid media, and the room nods. Nobody believes the figure.
I have sat in that meeting more than once. The number is the only thing on the page that looks like money, so it is the only thing anyone remembers a week later, and the assumptions behind it never made the slide.
The KPI List Is Not the Problem
You already know the list. Reach, sentiment, share of voice, message penetration, coverage volume. The AI Overview that greets anyone searching this term opens by defining PR measurement and pointing at AMEC's framework, three media monitoring vendors publish their own metric lists, and a row of PR agencies have reposted variations of the same list. Settled is the problem: a settled list never tells you which number is bigger than the other one.
The appetite is there. In Muck Rack's State of PR 2026, which surveyed 1,115 PR professionals and analysed 971 usable responses, 69% picked producing measurable results as how they prove PR's value to leadership or clients. A profession that agrees measurement is the answer and still argues about which number counts is short of a way to defend one, not short of metric names. Selecting the metrics is a different job from valuing what they measure, and the nine-KPI table in our guide to marketing KPIs already carries an owner, a cost and a source for every row.
Why Equivalent Ad Value Fails
AVE is the practice of taking what an ad would have cost in the same space and multiplying it by an editorial premium. The formula is published and easy to find: ad space or airtime, at the rate-card price, times a multiplier. Two of those three terms are guesses, and the third is a guess about a guess.
I understand why it survives. It turns an intangible into a currency, and procurement conversations run on currency. The lead who reports placements gets a polite nod. The lead who reports a figure gets a decision.
AMEC, the industry body behind the Barcelona Principles, put it plainly in version 4.0 in June 2025: "Invalid measures such as advertising value equivalents (AVEs) should not be used. Instead measure and evaluate the contribution of communication by its outcome and impact." That is not a survey result. It is the standard, and it needs no percentage behind it to be binding. A comparison metric is defensible to report. A hard result is not.
The Multiplier Has No Defensible Source
Ask what the multiplier should be and the published answers scatter. Releasd's explainer of the formula, read in October 2026, puts the range that circulates between 2x and 10x, and states that the industry never settled on a standard. So the number is set by whoever built the model, and two teams can look at identical coverage and report values five times apart.
The readership side is no better. Think Insights, reading Muck Rack's published estimator in October 2026, describes a method that starts from the outlet's monthly unique visitors. It then multiplies by 0.025, on the assumption that a share that size views any given article, and by $0.37 per visitor. The worked example in that report values a single article in a two million visitor outlet at $18,500. Two constants, both chosen. Change either one and the value of your coverage moves while the coverage itself sits exactly where it was.
Put the two halves together and you have a multiplier with no standard, applied to a readership share nobody measured for your article. The same placement comes out wildly apart depending on which end of each range you picked. Presscloud's AVE glossary makes the distinction that matters: this is a comparison metric, not revenue, and selling it as a hard result undermines every number that follows in the same report.
What to Substitute
Two things survive a follow-up question, and a third survives only if you describe it honestly. All three live on the upper half of the ladder AMEC's own framework describes, where outputs and outtakes give way to outcomes and impact.
Substitute | What it actually measures | What it survives | What it cannot see |
|---|---|---|---|
Cost per reached reader | What the same attention costs when you buy it at a documented CPC | A cost comparison against your own paid channel, with the benchmark's limits stated | Whether anyone read past the headline |
Branded search lift | Whether exposure caused more people to search for you | A controlled test with a p-value attached | What that search was worth |
Assisted conversion | Whether coverage appeared somewhere on a path that ended in a conversion | A pattern worth investigating | Which touch caused the conversion |
For the cost side there is a published anchor. LocaliQ publishes a 2026 benchmark for search advertising. It puts the all-industry average cost per click on the Google Search Network at $5.42, ranging from $1.63 to $9.87 across industries, built from 13,474 US search campaigns run between April 2025 and March 2026, and drawn from one company's client accounts rather than a Google-published industry average. Quote the number without that clause and you have repeated the AVE mistake with better sourcing.
What the CPC gives you is a denominator. Cost the reach you think a placement delivered at a rate somebody else documented, and you have a number that can be checked. It is not proof that your coverage reached that many people, so name the gap in the same sentence. The three substitutes also do not add up to one figure. Combine them into an index and you have rebuilt the multiplier with more steps.
Measuring Coverage Against Something Real
Branded Search, Used as an Instrument
Branded search is the cheapest honest proxy you have, and our pillar already treats branded search demand as a KPI in its own right, so I will not claim it as a discovery. What the pillar does not carry is the method.
Google's own Search Lift documentation, read in October 2026, describes how to run this properly. You split eligible users into an exposed group and a control group, then compare the difference in search behavior between them. A study can run for up to 28 days. Google documents a holdback between 1% and 50% of eligible traffic for user-based Conversion Lift, which is a separate product from Search Lift. The reporting includes a relative search lift figure and a p-value for incremental searches, which Google defines as the probability that the observed increase happened by chance.
Two things follow. The method comes from the ad platform itself rather than from a vendor with a dashboard to sell, and the limit keeps you honest: a lift study measures whether exposure caused more search, not what that search was worth.
Referral and Direct Traffic, Limits First
Referral traffic is the easiest number to pull and the easiest to over-read. Dark traffic arrives with no referrer at all, so it lands in direct and looks like brand demand. A placement in print or on broadcast sends you nothing measurable and may still have moved the business. The property you are reading from loses sessions of its own, and that side belongs to our guide to marketing analytics tools.
So referral traffic gives you a floor, not a total. A placement that sent some sessions sent at least that many, and it says nothing about the ones you cannot see. Direct is the worst evidence of the three, because it is the bucket everything unclassified falls into.
When a Series Beats a Single Report
A single campaign report answers a question about one campaign. That is the right artifact when the decision is already made. A series answers the question leadership asks: are we moving, and in which direction. You only get that from the same measurement, taken the same way, at the same interval. Share of voice enters the series as one input when you want competitive context, and the tooling behind it is set out in our guide to media monitoring tools.
The trap is treating a series as a bigger snapshot. A line going up on a metric that changed definition halfway through the year is two unrelated numbers joined by a straight line.
A Reporting Shape a CFO Will Read
One page. Three lines. No multiplier anywhere on it.
Line one is what you bought. The paid reach you are comparing against, the documented cost per click you costed it at, and one clause naming the benchmark's limitation. If you cannot name the source and the period, the number does not go on the page.
Line two is what you earned. Coverage volume, the covered readership estimate with its assumption stated, and branded search lift with the p-value beside it. If the p-value is weak, the line says so. A weak result reported honestly outlives a strong-looking one that collapses the first time somebody checks it.
Line three is what changed in the business. Conversions where coverage appeared somewhere on the path, stated as association rather than cause. If you need the models that separate those two, our guide to best attribution software owns that question.
Three lines is the amount of information a decision-maker can hold in their head at once. Assemble them from your own search, campaign and coverage data and the page refreshes instead of being rebuilt by hand, which is what the reporting surface in our marketing analytics feature is for.
The thing that makes this shape survive a review is that every number on it is smaller than the number you were reporting before. The covered readership estimate is smaller than the impressions total. That is the part nobody warns you about, and in an agency it is genuinely uncomfortable, because the old number was in last quarter's deck and the new one is smaller.
I have made this mistake myself. Early on I built a coverage report with a big number at the top because I assumed that is what the client wanted. The client asked one question about how it was calculated and the whole report stopped being useful, including the parts that were true. Since then the top of the report has been the smallest number on it.
Expect to be asked for the old number back. The honest answer is that the earlier figure was never measured, it was assumed twice. One more thing belongs on that page: a cheap placement that keeps getting cited by AI systems can out-earn an expensive one that nobody repeats.
Conclusion
PR measurement survives a CFO review when you stop pricing coverage as if it were media you bought, and start costing the comparison you can defend. The published CPC band gives you one side of it. A controlled search lift study gives you the other. Neither is a PR vendor's number, and that is why they hold up.
The question worth taking into your next review is not what your coverage was worth. It is which number on the slide you could still defend if somebody asked how it was calculated.
Frequently Asked Questions
How do you measure PR success?
Measure it on three levels and keep them apart. What was published: volume, placement, message presence. What that reached: covered readership, referral sessions as a floor. What changed: branded search lift, assisted conversions. Success is movement at the third level that traces back to the first.
What is a good PR ROI?
There is no universal number, and anyone quoting one is guessing about your cost base. A defensible version compares the documented cost per click against the covered readership you can support, and reports the gap rather than a ratio.
Is earned coverage worth what an agency charges for it?
Only if you measure it against what the same attention costs to buy. Ask for the covered readership estimate with its assumption named, the lift with its p-value, and referral sessions as a floor. An agency that can produce three numbers you can check is charging for something you can defend.
Build the Report That Survives the Question
The three lines on that page are assembled from search, campaign and coverage data you already have. In Allable they come from one workspace, so the number at the bottom of the slide is refreshed from source instead of rebuilt by hand each quarter.
