How to measure PR

Every quarter someone asks whether the PR is working, usually in exactly those words. The usual answer is a list of links, and a list of links is why PR is the first budget cut.

Measuring it properly needs the same fields on every piece and the same competitors every month.

Outputs and outcomes

Outputs are what ran, meaning the releases sent, pitches made and pieces of coverage. Outcomes are what changed because of it, such as your share of the conversation against competitors, the messages that made it into print and the buyer or AI assistant that now names you when asked.

Record both, but lead with outcomes, because outputs are your effort and the client is paying for the effect.

Most reports skip the layer between the two, which is the quality of each piece. Fourteen pieces of coverage tells a board nothing until it knows how many ran in titles it reads, how many were about you and how many just mentioned you, and what they said.

What to record on every piece

There are sixteen fields, and software can fill most of them the day the piece lands. Title, publication, journalist and date come first, and over time they build your history with each writer. Tier runs from 1 to 3 on a list agreed with the client. Format is one of news, feature, interview, opinion, review, round-up or case study. Then there's the tone towards you, and prominence, which is main subject, significant, minor or passing mention. Origin is press release, newsjack, journalist enquiry, interview, syndicated or organic. Add sector, medium and whether you secured it, and finally the four audience fields of domain authority, estimated views, reach and social shares.

Fill them in for the bad pieces too, because a log that only holds the wins can't show you what went wrong.

Tier, tone, prominence and the message

These four turn a count into a judgement. Tier says where it ran and tone how it treated you, while prominence separates a profile from a name-check. Message alignment tells you whether the piece carried what you wanted said, and it only works if you've written down three to five messages before the period starts.

Agree the tier list before the first report and don't move titles between tiers mid-year. Score depth, message alignment and favourability together and each piece has a quality score, so "our best piece this month" has a number behind it.

Estimated views, and why not AVE

Advertising value equivalence prices a piece at what an advert of the same size would have cost. On that measure a hostile piece is worth the same as a glowing one, and a passing mention the same as a profile. Online rate cards are notional, so nobody can check the figure. AMEC's Barcelona Principles, the industry's standard since 2010, say AVEs are not the value of communication, and most clients who still ask for one have never been offered anything better.

Offer estimated views instead. That's the audience a piece probably reached, modelled from the publication's traffic and the kind of page the article sits on, and labelled as an estimate. Show the publication's unique monthly visitors beside it as reach, under a column heading that says both are estimates. If a client insists on AVE, run both columns for a month and let them choose.

Share of voice, in the press and in AI answers

Share of voice is your slice of the coverage in a fixed set of publications and topics, against three to five named competitors, over a period. Count it by articles and again by audience. The two disagree more often than you'd expect. If you're trailing on count and leading on audience, your pieces ran in bigger titles, which is the better story.

Since buyers started asking ChatGPT for shortlists there's a second share of voice to report, which is how often an assistant names you against each competitor when asked your buying questions. Measure it the same way, with a fixed question set and the same competitors, run weekly. For each answer, record whether you were named, your position, your share of the names against competitors and the sources cited. The cited sources are where it joins back to PR, because they name the publications that earned the mention. What AI visibility is, and how to track it covers the method.

A monthly PR report template

The report has six sections, in this order, and they stay the same every month. The reader is a marketing director or a founder with ten minutes, so the summary has to stand on its own.

Sections 5 and 6 should talk to each other. A title the assistants cite for your category that hasn't covered you yet is next month's pitch.

The cadence

Log within 48 hours of publication, with a screenshot, because headlines change and paywalls drop. Read the week's coverage against the plan on Friday. Report in the first week of the month. Judge targets quarterly against the same quarter last year, and review the tier list and the question set then, because share of voice and AI visibility are trend measures and need three months before anyone claims a direction.

Agentcy's coverage reports record the sixteen fields and the quality score on each piece as it is found, and its AI visibility tracker records the four AI measures weekly, so you aren't building the report from a spreadsheet. It doesn't find print or broadcast pieces.