How to Measure PR: The Numbers That Survive a CFO's Questions

PR measurement is the practice of connecting communications activity to outcomes that a budget holder cares about. The numbers that hold up are tied to stated objectives, defined the same way over time and honest about what coverage did and did not cause.

Every communications team eventually has to answer a version of the same question from someone holding the budget: what did this actually do. The honest reporting practices in this piece are the ones that survive that question being asked twice, by someone who is genuinely trying to understand the number rather than being handed a report to file.

Start with what you said you would do

The Barcelona Principles, the measurement standard maintained by AMEC and refreshed to version 4.0 in June 2025, open with a sentence that reads like boilerplate and is not: “Setting clear, measurable objectives is a critical prerequisite for effective communication planning, measurement and evaluation.”

The word doing the work there is prerequisite. Not best practice, not a recommendation. A prerequisite, meaning the rest of it does not function without this.

Most PR reports are difficult to defend because there was never anything to defend them against. The work was commissioned as “get us coverage,” which has no failure condition, so the report at the end can only describe what happened and hope it sounds like a lot. A team that agreed in advance that the objective was to move trade-press framing away from price and towards supply reliability, in three named publications, across one quarter, has a question with an answer. It can also fail, which is exactly what makes the success mean something.

The practical test: before the work started, could you have written down what a disappointing result would look like? If not, whatever appears in the report at the end is a description rather than a measurement, and the person reading it will feel that even if they cannot name why.

Four rungs, and why most reports never leave the first

AMEC’s Integrated Evaluation Framework separates what communication produces into a sequence, and that sequence is the most useful single idea in PR measurement, because it explains in one move why so many reports feel unsatisfying to the person receiving them.

Outputs are what you produced and placed. Articles, posts, briefings, the coverage itself. Entirely within your control, and entirely about you.

Outtakes are what the audience took from it. Whether they noticed, understood, remembered, believed. This is the first rung on which the audience appears at all.

Outcomes are what the audience then did or felt differently. Changed opinion, increased consideration, a shift in stated preference, an enquiry that would not otherwise have been made.

Impact is what that meant for the organisation. Revenue, retention, recruitment, a regulator who moved slowly rather than quickly.

Principle 6 of version 4.0 is explicit that reporting should span the range: “Measurement and Evaluation should report outputs, outcomes, and impact related to the organization and stakeholder audiences.”

Almost every PR report in circulation reports the first rung exclusively, and then presents it in language borrowed from the fourth. That is the real defect in AVE and reach, and it is worse than inaccuracy. They are output measures dressed as impact measures, and the mismatch is what a finance director senses immediately without necessarily being able to articulate it. A number describing how much you published, denominated in currency, invites a question about return that the number is structurally incapable of answering.

The reframe that makes all of this easier: stop trying to make outputs sound like impact. Report outputs as outputs, honestly and briefly, then spend the rest of the report on the rungs above them.

Retire these first

Advertising Value Equivalency. The premise, that an article is worth what an equivalent ad space would have cost, multiplied by some factor for earned credibility, has never survived contact with an actual finance team asking where the multiplier came from. It varies by agency, has no external standard, and conflates two things, paid reach and earned trust, that behave completely differently in a reader’s mind. If a report still contains an AVE number, it is the first thing to cut, not the last. The longer version of this argument, including what AMEC changed in June 2025 and why fifteen years of calling AVE invalid did not shift anything, is here.

Raw reach and impressions. A number built from a publication’s self-reported audience figures, summed across every mention regardless of whether anyone actually saw or read the piece. It inflates easily, cannot be audited, and answers a question nobody in finance actually asked, which was never “how many eyeballs theoretically passed near this” but “did this change anything.”

Clip counts. The number of articles published about you is a volume metric, not a performance metric. Ten mediocre mentions are not better than three that shifted the conversation, and a report led with a clip count trains its audience to expect volume as the measure of success, which sets comms up to be judged on the wrong axis every quarter after.

What actually holds up

Share of voice, defined narrowly and honestly. What percentage of coverage in your category mentions you versus a specific, named set of competitors, in a specific time window. This survives scrutiny because it is comparative rather than absolute, it is auditable if someone wants to check the underlying mentions, and it answers a question a CFO actually cares about: are we gaining or losing ground against the companies we compete with for the same customers and the same narrative.

Sentiment trend against your own baseline. Not sentiment as a static percentage, which means little without context, but sentiment moving up or down relative to what is normal for your brand specifically. A brand that runs 65% positive most quarters and drops to 55% has a real signal worth explaining. A brand that always runs around 40% positive because its category is inherently contentious is not having a crisis at 38%, and reporting it as one erodes credibility the first time someone checks the history.

Message pull-through. Of the coverage that ran, how much of it actually carried the point the team was trying to land, versus coverage that mentioned the brand but told a different story entirely. This is more work to track than the other numbers, because it requires someone to read the coverage rather than run a keyword count, but it is the number that most directly answers whether the PR effort did its actual job.

Correlation with a business signal, stated honestly. Search demand, site traffic, or category-specific interest metrics tracked in the same window as major coverage moments. The discipline here is language: showing that a spike in searches for the brand name coincided with a major placement is a legitimate and useful observation. Claiming the placement caused a specific number of leads or a specific revenue figure, without the tracking infrastructure to actually support that claim, is the fastest way to lose the room’s trust the next time a number is presented.

Climbing past outputs without a research budget

The standard objection to the framework above is that outtakes and outcomes require survey research, and most teams have no budget for a tracking study. That is partly true and it is not a reason to stay on the first rung, because several outtake and outcome signals are already sitting in systems the business is paying for.

Branded search volume. Someone who searches your brand name after encountering coverage has demonstrably taken something from it. This is one of the cleanest outtake proxies available, it costs nothing, and it is time-stamped finely enough to line up against a specific placement date.

The shape of direct and organic traffic, not the total. A placement that mattered leaves a visible step. One that did not leaves nothing. The absence is as informative as the presence, and reporting both is precisely what makes the presence credible when you do report it.

Language showing up in sales and support calls. If the commercial team starts hearing a phrase that originated in your coverage, that is an outtake with a human being attached to it. Capturing this is unglamorous and requires persuading people to write things down, and it tends to produce the most persuasive single item in the whole report, because it is the one thing a sceptical executive cannot dismiss as a coincidence in the numbers.

Unprompted inbound of any kind. Speaking invitations, analyst enquiries, partnership approaches, candidates who name the coverage in an application. Each one is a small, dateable outcome, and a running log of them across a year is more convincing than any chart.

A short survey of one audience, once a quarter. Cheaper than most teams assume when it runs against an existing customer or subscriber list rather than a purchased panel. It will not be nationally representative. Saying so plainly costs nothing, and it still moves the conversation from “we got coverage” to “here is what the people we actually care about said afterwards.”

None of these is proof. All of them are evidence, which is a different and considerably more achievable standard.

Building a report that survives being asked about twice

The test for any PR report is not whether it looks complete in the room. It is whether the person who received it can bring it back a week later, ask a follow-up question about any number on the page, and get an answer that holds up. Most reports fail this test not because the underlying work was bad, but because the reporting reached for numbers that were easy to produce rather than numbers that were defensible.

A report built around the four categories above is shorter than most current PR reports, and that is the point. Share of voice against named competitors. Sentiment trend against baseline, with the baseline stated. A short, specific list of the coverage that mattered that period and a sentence on why. One or two business signals tracked honestly, without overclaiming causation.

This format also forces a useful discipline earlier in the process: a team that knows it will have to defend message pull-through at the end of the quarter pays more attention to what message it is actually trying to land at the start of it, rather than treating any coverage as a win regardless of what it said.

The number you cannot produce, and how to say so

Every honest measurement practice eventually arrives at a question it cannot answer. Did the coverage cause the quarter. Usually there is no way to know, because the campaign ran alongside a price change, a competitor’s stumble, a seasonal pattern and a paid media flight, and no dashboard will separate those.

Version 4.0 of the Barcelona Principles gives this situation its own principle: “Ethics, governance and transparency with data, methodologies and technology builds trust and drives learning.” In practice that comes down to three habits.

State the method next to the number, not in an appendix. One sentence saying share of voice covers these four competitors, these source types, this window, is worth more to a sceptical reader than another chart. It also makes the number harder to quietly change later, which is the actual point.

Report the movement you cannot explain. A quarter where sentiment improved and nobody knows why is a genuine finding. Saying so costs nothing the first time and buys a great deal of credibility by the fifth, because it establishes that the report is describing reality rather than constructing a case.

Separate correlation from claim in the sentence itself. “Branded search rose 34% in the week of the placement” is defensible. “The placement drove a 34% rise” is an assertion about causation you cannot support. The difference is one verb, and it is the verb a CFO will pick up on.

The instinct is that admitting uncertainty weakens a report. In practice the opposite happens. A report that names its own limits is the only kind that gets believed about the things it does assert, because the audience has watched it decline to overclaim at least once.

Where this gets harder, and why it is still worth doing

Measuring PR honestly is harder than measuring it with AVE and clip counts, because the honest numbers require more infrastructure: consistent sentiment scoring across a real time series, a defined competitor set tracked the same way every period, and someone willing to read coverage rather than just count it.

The payoff is a report that survives being questioned, which changes the relationship between comms and the rest of the business. A team whose numbers hold up under scrutiny gets asked fewer skeptical questions over time, not more, because the report has earned a track record of being straight rather than flattering. That is a slower thing to build than an impressive-looking AVE total, and it is the only version of PR measurement that keeps its credibility past the first quarter someone checks the math.

Common questions

How do you measure PR performance?

By tracking a small set of numbers that hold up under scrutiny: share of voice against named competitors, sentiment trend against your own baseline, message pull-through in the coverage you actually earned, and a documented link between coverage moments and business signals like site traffic or search demand. Reach and AVE do not survive this bar.

What is the best KPI for PR?

There is no single best KPI, because PR does several different jobs. The honest approach is a small dashboard: share of voice for competitive position, sentiment trend for reputation health, message pull-through for whether the story landed as intended, and a qualitative log of the coverage that actually mattered that quarter.

Why is AVE (advertising value equivalency) considered a bad metric?

AVE treats an article as if it were an ad bought at the same size, which assumes readers trust earned coverage and paid advertising equally. They do not, the multiplier used to inflate the number is arbitrary and varies by agency, and it cannot be defended under any real scrutiny, which is why most serious measurement bodies have recommended dropping it for over a decade.

How do you connect PR to revenue?

Rarely directly, and claiming otherwise is usually the first sign of a report that will not survive questioning. What works is showing correlation: search demand or site traffic moving in the same window as a major placement, share of voice trending against a competitor whose pricing or share also moved, without overstating the certainty of the link.

What is the difference between outputs, outtakes, outcomes and impact in PR measurement?

They are the four stages of AMEC's Integrated Evaluation Framework. Outputs are what you produced and placed, such as articles and posts. Outtakes are what the audience took from it, such as awareness or recall. Outcomes are what they then did or felt differently. Impact is what that meant for the organisation. Most PR reports only measure outputs and then describe them in the language of impact, which is the root of why they fail under questioning.

What should a monthly PR report actually contain?

Fewer numbers than most reports currently have. Share of voice, sentiment trend, a short list of the coverage that mattered and why, and one or two business signals tracked in the same window. Everything else is usually there to fill space rather than answer a question anyone asked.