Every PR engagement eventually arrives at the same conversation. Someone in finance asks what the retainer returned, and the agency produces a report full of impressions, reach, and advertising value equivalency. The numbers are large, the numbers are meaningless, and everyone in the room knows it.
The honest position is that PR is partially measurable. Some of its effects can be tracked with reasonable confidence, some can be inferred with caveats, and some cannot be measured at all without lying. A framework that admits the third category is more useful than one that pretends everything reduces to a single figure, because the pretense is what destroys credibility when the CFO starts pulling on threads.
Why PR resists attribution
Three structural properties make PR harder to measure than paid channels, and they are worth naming because they are not going away.
Coverage has no click. A reader encounters an article, forms an impression, and acts weeks later through a channel that gets the credit. The article is upstream of a search, a referral conversation, or a direct visit, and last-touch attribution assigns the value to whatever happened last.
Coverage compounds and decays unevenly. A feature published in March is still working in October, still being found in search, still being sent by one person to another. A campaign burst has a long tail that no monthly reporting window captures.
Coverage changes conversion rather than volume. Its most valuable effect is frequently that deals already in motion close faster and at better terms, which shows up in sales metrics that nobody is attributing to communications.
None of that makes PR unmeasurable. It makes single-number ROI the wrong instrument.
Four signals worth tracking
These four are trackable with ordinary tooling and, taken together, describe the direction of travel honestly.
Referral traffic from the publication and its downstream. The simplest signal. When an article runs, some readers click. Track sessions from the publication’s domain and from the syndication and aggregation that follow, and track what those sessions do. Volumes are usually modest and that is expected. A tier-1 feature that sends a few hundred highly qualified sessions is doing its job; the number is small because the value is not in the click.
Branded search volume. The most reliable proxy in the set, and the one most teams ignore. People who encounter you in an article often do not click. They search your name later. Pull branded query impressions and clicks from Search Console, mark the publication dates on the timeline, and look for step changes. This captures the delayed, indirect response that referral traffic misses. It requires a clean before-and-after window to be worth anything.
Close-rate and cycle-length deltas. The most valuable signal and the one requiring actual cooperation from sales. Add one question to the intake process: had this prospect encountered coverage of you before the first call. Then compare close rate, cycle length, and average deal size between the two groups. This is where the real economics of editorial coverage usually live, and it is the number that ends the ROI argument in either direction.
AI citation share. Newer and cruder, but increasingly the surface where buyers form first impressions. Maintain a fixed list of the questions a prospective client would ask an assistant, run them on a schedule, and record whether you appear and what gets cited. It is manual and imprecise, and it is still the best available read on a channel that is now upstream of a meaningful share of consideration. The mechanics of why coverage drives those citations are in press coverage and AI search visibility.
The discipline that makes any of it work
None of the four means anything without a baseline, and the baseline is where most measurement programs fail before they start.
Record the numbers for the ninety days before the engagement begins. Branded search volume, direct traffic, inbound lead volume, close rate, average cycle length, and the assistant-citation snapshot. Freeze them. Do this before the first pitch goes out, because you cannot reconstruct a baseline afterward and every attempt to do so is contaminated.
Then hold the comparison honestly. If you launched a paid campaign in the same window, you cannot attribute the lift to PR, and saying so is what makes the rest of your reporting credible.
What cannot be measured honestly
A framework is only trustworthy if it marks its own limits.
Reputation cannot be measured. The effect of a partner being the named authority in a field on a general counsel’s shortlist is real, consequential, and not available as a number.
Deals that never happened cannot be measured. Coverage that prevented a prospect from dismissing you before the first call is invisible by construction, and it is plausibly a large share of the value.
Recruiting effects can be observed but rarely isolated. Candidates cite coverage in interviews. Whether the coverage produced the candidate or merely confirmed a decision already forming is not knowable.
Long-tail compounding cannot be attributed within a reporting period. An article that produces an inbound inquiry two years later will not be credited to the engagement that produced it, and no attribution model will fix that.
The correct treatment of these is to name them explicitly as unmeasured value, once, and then not to smuggle them back in as an inflated number. Clients respect the distinction. They do not respect a report that claims everything.
Metrics to refuse
Advertising value equivalency. AVE takes the physical space your coverage occupied and prices it as if you had bought an ad there. The industry’s own measurement standards have rejected it for years, for the obvious reason that editorial coverage and advertising are not equivalent goods, which is the entire premise of buying PR in the first place. Any report leading with AVE is a report designed to produce a large number.
Impressions and potential reach. Usually sitewide monthly traffic multiplied across every domain the piece appeared on, including syndicated duplicates. It measures the theoretical audience of the websites, not the readership of your article. We covered how this inflation is constructed in how to choose a PR firm.
Placement counts without surface labels. Twelve placements means nothing until you know how many were staff-reported and how many were the same release republished.
How long before any of it shows up
Impatience is the most common cause of an engagement being judged a failure while it is still working. The signals do not arrive together, and knowing the order prevents the wrong conclusion at the wrong moment.
Referral traffic arrives first and is the least meaningful. It spikes on publication day, decays within roughly a week, and tells you almost nothing about value. Teams that watch this number closely tend to conclude too early that nothing happened.
Branded search moves next, usually over weeks rather than days, and it moves as a step rather than a spike. A single placement rarely produces a visible change. A run of coverage over a quarter generally does, which is why a one-placement engagement is nearly impossible to evaluate on this metric.
Sales-side effects take a full sales cycle to become visible, by definition. If your cycle runs four months, you cannot assess close-rate deltas at ninety days. This is the most common structural mismatch between reporting cadence and reality, and it is worth stating at the start so nobody is surprised.
Citation share is the slowest and the noisiest. Systems re-crawl on their own schedules and answers vary between runs for reasons that have nothing to do with you. Look at it quarterly, not monthly, and look for direction rather than movement.
The practical consequence is that a fair evaluation window is one full sales cycle plus a quarter, and anyone promising a verdict sooner is describing referral traffic.
What the range actually looks like
Buyers reasonably want to know what the market charges before they can judge whether a return is plausible. Published industry ranges for retained editorial PR span a wide band, from low four figures monthly for narrow, single-market scopes to well into five figures monthly for national programs with multiple named targets and a dedicated team. Project-based single-placement work occupies its own band, generally lower and more variable.
Two things are worth drawing from that spread rather than from any single number in it.
The first is that the range is wide because the work is not one thing. A scope with one named target, one spokesperson, and one market is a fundamentally different labor cost than a program running multiple beats across several publications with a compliance layer. Comparing quotes without normalizing the scope produces a meaningless comparison.
The second is that pricing far below the band for the outcome described is a signal rather than a bargain. Earned work carries an irreducible labor cost: research, angle development, relationship maintenance, and pitching that mostly fails before it succeeds. A quote that does not cover that labor is usually funding distribution instead, which returns to the measurement problem this post opened with.
Because scope drives the number so heavily, we price by engagement rather than publishing a rate card, and the scope is defined before the number is. The first conversation is about what the job actually is.
A working scorecard
For most engagements, a defensible quarterly report has five parts and fits on a page.
The list of articles, each labeled staff-reported, contributed, or syndicated, with links. Referral sessions from those articles and their downstream, with what those sessions did. Branded search volume against the frozen baseline, with publication dates marked. Sales-side deltas for prospects who had prior exposure to coverage versus those who had not. The assistant-citation snapshot against the baseline snapshot.
Then a short paragraph naming what is not in the report and why. That paragraph is what makes the rest of it believable.
So is PR worth it
It depends on a question most buyers have not asked themselves: what is the actual job.
If the job is direct-response volume, PR is usually the wrong instrument and paid acquisition will outperform it on cost per acquisition. Buying PR to do a performance channel’s job produces exactly the disappointment you would expect.
If the job is credibility that shortens sales cycles, survives diligence, supports premium pricing, and determines how you are described when someone asks a machine about you, then the alternatives are worse and mostly unavailable at any price. You cannot buy a journalist’s independent assessment, which is precisely why it is worth something.
The engagements that go badly are almost always ones where those two jobs were confused at the point of sale. The ones that go well start with a baseline, a named target, and an agreement about what will and will not be claimed.
The PR Summit Editorial writes for founders, partners, and principals on the editorial work behind tier-1 coverage.