Messung
Measuring PR that actually matters
You ran a PR push. Coverage landed. Then someone asks the only question that matters: did it work? Most teams reach for the nearest big number. A clip count. An impressions total. An ad value equivalent that turns a mention into a dollar figure. The numbers look like proof. They are theatre.
This guide is for founders and lean marketers who do their own PR and need an honest answer, not a flattering one. The gap between those two answers is where budgets get wasted and good programs get killed for the wrong reasons.
The argument runs through the whole book. A metric earns its place only when it connects coverage to something the business cares about: awareness, consideration, pipeline, and now whether AI assistants cite you when a buyer asks who to trust. Everything else is decoration.
You will not get a formula that proves PR caused a sale. Nobody honest can sell you that. You will get a way to measure that respects the limits of what you can know, triangulates the signals you can actually see, and still tells you enough to decide what to do next month. Every method here arrives with what it can show and what it cannot. Then we build the monthly report and the learning loop that turn measurement into sharper pitches and more coverage that counts.
Chapter 1
Big numbers are comforting. After a launch you want to point at something large and say, that worked. The PR industry is happy to supply the number, and it is almost always the wrong one. Clip counts, impressions, and ad value equivalent share a single flaw. They measure the act of publishing, not any effect on a buyer. Each one can climb all month while your business sits perfectly still, and that gap is the whole problem.
These metrics survive because they flatter everyone at once. The agency looks productive, the marketer looks effective, and the founder gets a large figure to forward to the board. They are also fast to produce, which matters when a report is due and the week got away from you. Flattering and fast is a hard habit to break. It breaks only when you replace it with a smaller number that answers a question you actually have.
A clip count is the number of pieces that mentioned you. Fifteen hits sounds better than three, so fifteen goes in the deck. The count treats a one-line mention in a dead roundup the same as a feature in the single publication your buyers actually read. One sharp feature in the outlet your market forwards to each other will do more than fifty mentions nobody sees, yet the count scores the fifty higher. Volume is cheap: one wire release can spawn a dozen syndicated copies no decision-maker will ever open. Worse, chasing the count pushes you toward the mass blast that irritates the few journalists who matter, which costs you the coverage you actually wanted. A high clip count tells you someone was busy. It does not tell you one person thought differently about you afterward.
Impressions, often relabeled potential reach, usually start from an outlet's entire audience and assume every one of those people saw your story. None of them did, or at least you have no way to know which ones did. The figure ignores who read past the headline, whether your name survived the first paragraph, and whether the reader was a buyer, a competitor, or a bot. Even the people who genuinely read a piece mostly do nothing about it, which the number also cannot see. Add impressions across every clip and you get a total in the millions that feels like scale and guides no decision, and you cannot even compare it honestly against last quarter. A measurement you cannot act on is not a measurement. It is a mood set in a large font.
Ad value equivalent, or AVE, takes the space your coverage occupied, prices it as though you had bought an ad in that spot, and sometimes multiplies the result by an invented factor because editorial supposedly outperforms advertising. The method fails at the root. It values a story by what an ad would have cost, which has nothing to do with whether the story moved anyone. A long, dull feature in a trade journal your buyers ignore will out-score a three-line mention in the one newsletter they trust and forward, purely because the feature filled more column inches at a higher rate. It can even score a damaging article as a win because the article was long. This is not a fringe complaint. AMEC, the industry's own measurement association, wrote the Barcelona Principles specifically to retire AVE, and states plainly that ad value equivalents do not represent the value of communications. When the trade that once profited from a metric votes to abandon it, you can drop it with a clear conscience.
A bad metric is not harmless, because what you measure quietly steers what you do. Reward clip counts and your outreach turns into a spray of generic releases, since quantity is the fastest way to move the number. Reward impressions and you chase the biggest outlet in the room regardless of whether its readers are your buyers. Reward AVE and you optimize for long articles in expensive publications, even when a short mention in the niche newsletter your market trusts would do far more work. The number you celebrate becomes the work you produce. Choose numbers that reward relevance and real demand, and your pitching sharpens on its own.
Before a metric goes in your report, put it through three questions. Does it connect to a business outcome you care about, such as demand, pipeline, or being cited where buyers ask for recommendations? If the line from the number to the outcome needs three leaps of faith, drop it. Would it change a decision? If the figure doubled or halved and you would do nothing differently, it is scenery. Can you collect it honestly and repeatedly, the same way every month, without a vendor massaging the inputs? A modest number you can reproduce beats an impressive one you cannot. A quick way to apply this: for every metric on your current report, write down the decision it would change, and if the space stays blank, the metric comes off the page.
Run your dashboard through those questions tonight. Most of what survives will be small, specific, and a little boring. That is the point. Honest measurement trades the rush of a seven-figure impressions total for a short list of signals that tell you whether to keep going and where to push next. The following chapters build that list, starting with the outcomes every signal has to serve.
Takeaway: If a number cannot change a decision, it is decoration, and AVE is the first decoration to throw out.
Chapter 2
Measurement starts at the wrong end when you begin with the data you happen to have. Begin instead with the business result you need, then work backward to the smallest set of signals that would tell you whether you are getting closer. For most founders the result is some version of qualified demand: the right people arriving, already warm, ready to talk. PR rarely closes that demand on its own. It works upstream, shaping who knows you and who trusts you long before a sales conversation ever starts.
Think in three stages the business already understands. Awareness: do the right people know you exist? Consideration: when they have a need, do you make their shortlist? Pipeline: do those people turn into conversations and deals? PR touches all three, hardest at the top and more faintly as you move down, because by the time a deal closes many other things have happened. A feature can plant your name in a buyer's head a year before they are ready, then a comparison article nudges you onto the shortlist, then a founder quote in a trusted outlet gives a champion the cover to pick you. Measuring PR well means watching the stages it moves first and resisting the pressure to make it prove revenue by Friday.
There is now a stage sitting above awareness. Buyers ask AI assistants who to trust before they ever run a search or open a website. They type a question into ChatGPT, Claude, Perplexity, or Gemini and read the three names that come back. Often the assistant answers in full and the buyer never clicks through to anyone, so the only trace you get is whether your name was in the answer at all. That makes the AI layer both the earliest stage and the least visible, and the one most likely to be missed if you only watch website analytics. If you are not among the names an assistant returns, you are invisible before the funnel even begins, and no amount of downstream optimization fixes it. So AI citations join the outcomes that matter: when a buyer asks for the best options in your category, are you cited, and how often compared with your competitors? This trust is earned the way trust has always been earned, through credible coverage and clear positioning, and now it surfaces in a new place you have to watch.
Each stage has a signal you can actually watch:
Say the result you need this quarter is qualified demo requests. Work backward from there. The demo request is the pipeline signal, so you watch how many arrive and how many mention where they first heard of you. One stage up, consideration, you watch branded searches that pair your name with pricing or a rival, because those are people comparing you on purpose. Above that, awareness, you watch branded search volume and direct traffic for the plain fact of more people knowing you exist. At the very top you watch whether an assistant names you when someone asks for the best option in your category and city. Five signals, one per stage plus the AI layer, each tied by a visible thread to the demo request you care about. That thread is the test. If you cannot draw it, the signal does not belong on the page.
The mistake now is to track all of these at once and drown. Choose the one business result you are chasing for the next two quarters. Trace it back through the stages and pick one or two signals per stage, the ones you can read honestly and repeatedly. Four to six signals is plenty. A short dashboard you check every month beats a vast one you abandon by March. Hold these signals for what they are: evidence that PR is pushing the business the right way, not a ledger that assigns each dollar to a headline. Awareness and trust move before money does, often by months, and they move alongside everything else you are doing. Name the few that matter, watch them over time, and you will know whether your PR is working long before the revenue line could tell you. Review the list every quarter and prune it. A signal that stopped telling you anything new has earned its way off the dashboard, and a new business goal may call for a different one. The point is a living short list, kept honest, not a monument you built once and stopped reading. The next chapter turns these signals into an attribution model you can run yourself with no budget.
Takeaway: Start from the business result, work backward, and watch four to six honest signals instead of everything you can count.
Chapter 3
There is no single source of truth for PR attribution. Anyone who promises one is selling the AVE of this decade. What you can build instead is a composite: four cheap reads that each catch a different slice of reality, none reliable alone, useful together. You can run all four with tools you already pay nothing for. Think of it as triangulation, the way a sailor fixes a position from several rough bearings instead of one exact instrument nobody handed you. Here is each one, with what it shows and where it lies.
Add one question to your signup, contact, and demo forms: how did you hear about us? Ask it again, in plain words, on sales calls. This is the only method that reaches offline and untracked influence, the podcast someone half-remembers or the article a colleague forwarded. Phrase it as an open question and resist a tidy dropdown of channels, because a dropdown teaches people the answer you expect and buries the ones you did not. On calls, ask it as a story: what first put us on your radar? The stories are messy and more honest than any menu. What it shows: the touch the buyer consciously credits. Where it lies: people forget, they name the last thing they remember, and PR often primes a buyer who then credits the search that followed. Only people who converted ever answer, so you never see who you lost. Tag the free-text answers into a few buckets yourself at month end, so the mix stays readable without forcing buyers into your categories at the door. Read the mix over months, as a direction, not a tally of any single week.
When coverage works, more people go looking for you by name. Google Trends shows the shape of public interest in your name; Search Console shows the clicks and impressions your own site earns for brand queries. Use Trends to see whether curiosity rose and Search Console to confirm it reached you. What it shows: rising intent, which is what PR tends to create. Where it lies: it rarely proves which placement caused the bump, it is noisy for small brands where a dozen searches swing the line, and seasonality and your other marketing both bleed into it. The fix is discipline: know your normal before you read the spike. Set a baseline for the weeks before a push, then mark the placement date on the chart and read the lift against that.
Some readers click. Referral traffic in your analytics shows visits sent from a publication's domain, and UTM tags on any link you control, your own posts, your newsletter, and partner placements, let you label those visits precisely. Tag only the links you actually own, because a UTM you invent for a journalist's article will not survive to publication. What it shows: the clickthrough you can actually see. Where it lies: most readers never click, many outlets strip or nofollow links or keep the traffic inside their own walls, and print, broadcast, and AI answers usually pass no referral at all. A strong placement can send almost no referral traffic and still move your whole market. The honest version of this read is narrow and reliable: it counts the clicks you can see and stays silent about the far larger number you cannot.
Direct traffic is the junk drawer of analytics: people typing your address, untagged links, and a pile of unknowable sources. Watch it anyway. A jump in direct traffic in the days around a placement is a shadow of influence the other methods missed, usually people who saw your name somewhere and came straight to you later. Treat the jump as a question, not an answer. What it shows: a rough proxy for untracked reach. Where it lies: it is correlation at best, it absorbs all kinds of unrelated activity, and it can never prove a cause by itself.
Put the four together and the method becomes usable. Keep a simple timeline of your analytics and annotate it with every placement date. Hold a baseline for branded search and direct traffic so you know what a normal week looks like. When a placement lands, read all four around that date. If self-reported mentions of the outlet rise, branded search lifts, referral arrives, and direct traffic jumps together, you have a real signal. If only one of them twitches, stay honest and wait. Resist over-reading one dramatic week; the composite earns its trust across several placements, as a pattern that repeats. No single line is proof. The agreement across all four, seen again and again, is as close to proof as PR will ever give you.
Often the reads disagree, and disagreement is information. A placement in a high-trust outlet that passes no link will show up as branded search and direct traffic with almost no referral, which tells you it worked offline. A placement that drives a burst of clicks and no lasting search interest reached people who were curious for a moment and gone by the weekend. Read the pattern, name what it implies for that outlet, and carry it into next month. The goal is never one verdict for all coverage; it is a read per placement that teaches you which kinds of outlets move which signals for your brand.
Takeaway: Run four cheap reads together and trust the pattern they agree on, never a single line on its own.
Chapter 4
Coverage ran on Tuesday. Traffic jumped on Wednesday. The temptation is to draw a straight line between the two and claim the win. Resist it. That straight line is where PR measurement stops being honest, and once a founder starts believing flattering lines, every decision downstream tilts with them. This chapter is about reading your own results without fooling yourself.
A spike after coverage is a reason to look closer, nothing more. Ask what else happened that week. Did you turn on ads, ship a launch, send a big newsletter, or catch a seasonal bump? Did a second outlet pick up the story, or did a post travel on its own? Every one of these moves the same numbers PR moves. Until you have ruled them out, the coverage is one candidate explanation among several. Treating correlation as proof is the exact error that let AVE survive for decades, dressed up as rigor. The more a result flatters you, the harder you should look for the other cause, and the standard you apply to your own win should match the one you would demand of a competitor claiming it.
Here is how to read a jump without lying about it. Put every spike against its baseline first. A forty percent rise in branded search sounds decisive until you notice the line swings thirty percent week to week on its own, at which point forty percent is barely a ripple. The baseline is what separates a signal from the normal churn of small numbers, and small brands live in small numbers. Then lay the four reads from the last chapter over the placement date. If branded search, direct traffic, referral, and self-reported answers all move together, and nothing else large ran at the same time, you have a defensible case that the coverage did real work. Write it down with a confidence level and the rival explanations you considered and rejected. If only one read moved, or you also launched a campaign that week, say so and hold the claim loosely. The goal is a verdict you would still defend if a skeptical investor pulled on the thread.
Separate the signals that move first from the ones that move last. Coverage quality, AI citations, and branded search are leading indicators: they respond within days or weeks of good PR. Pipeline and revenue are lagging indicators: they arrive quarters later, after many other touches. The timing mismatch is where founders go wrong. If you judge this month's coverage only by this month's revenue, you will kill programs that were working and keep ones that happened to coincide with a good quarter. Judge PR mainly on its leading indicators, track the lagging ones over time, and watch whether the leading signals you can move really are the ones that precede the revenue you want. A fair rule of thumb: give a placement a full quarter before you expect it in the pipeline numbers, and judge it in the meantime on whether it moved awareness and citations. When they stop predicting it, change what you measure.
A closed deal usually carries many touches: an article, a search, a referral, three emails, a demo, a case study. PR is often the earliest of these and the easiest to forget by the time the deal signs. First-touch attribution hands PR the entire deal and flatters it. Last-touch attribution hands the deal to the final email and erases PR completely. Both lie, in opposite directions. The honest move is to record that PR appeared in the story, as an assist, and stop short of claiming the whole value. If you have the scale for it, a real test beats any attribution model: hold PR out of one region, stagger launches across markets, and compare. Most lean teams do not have that volume, and pretending a noisy before-and-after is a controlled test is just a quieter form of the same dishonesty. When you cannot run the test, say what you believe and how sure you are, and leave it there.
A quick warning about tools. Plenty of dashboards will hand you a tidy percentage for PR's contribution to pipeline, down to the decimal. Distrust any model that never admits it does not know. Real attribution is full of gaps, and a number that hides the gaps does more damage than an honest shrug, because it invites you to bet on a precision that was never there. The best tools in this space are the ones that show their working and leave the conclusion to you. Use them to gather the reads faster, and keep the judgment, and the confidence level, with a person who has to defend it out loud.
Make honesty a habit instead of a mood. Before a push, write down what you expect to move and by roughly how much. After it, compare what actually happened against that prediction and against the other things that ran. Keep a confidence level on every claim in place of a flat yes or no. Over a few cycles this record becomes more valuable than any single report, because it shows which of your signals tend to lead to pipeline and which were noise you once got excited about. That memory is what keeps the next chapter's report honest.
Takeaway: A spike is a hypothesis, so triangulate it, state your confidence, and claim only the credit you could defend out loud.
Chapter 5
Most PR reports are a clip dump with a total stapled to the top. Pages of logos, a count of hits, an impressions figure in the millions, maybe an AVE in dollars. The reader skims it, feels vaguely positive, and decides nothing. A report that drives no decision wastes the hours spent making it. In this era the report has a sharper job: show what moved, what it means, and what you will do next, on a single page.
A PR report speaks to someone deciding whether to keep funding the work. That reader needs three things: a clear read on whether the program is working, enough honesty to trust the read, and a plan for next month. Everything on the page should serve one of those three. If a section exists only to look busy, cut it. Keep it to one page, because a page gets read and a slide deck gets skimmed.
A report that shows movement has a stable shape:
The order carries the logic. Start with the verdict so a busy reader can stop after one line. Lead the evidence with coverage quality and AI visibility, because those are the things PR moves first and most directly. Put demand and pipeline next, honestly hedged, so nobody mistakes a correlation for a guarantee. End with the learnings, which turn the report from a record into a plan.
Leave off AVE entirely. Leave off raw impressions as a headline figure. Leave off the exhaustive clip list; if someone truly wants every mention, link it as an appendix and keep it off the page. These cuts are not cosmetic. Every line you remove that could not change a decision makes the lines that matter easier to see and easier to act on.
The report's power compounds when its shape holds still. Use the same sections, the same signals, and the same baselines each month so the reader sees movement at a glance instead of relearning the format. A report that rearranges itself to fit the month's good news is quietly dishonest, because it hides the comparison that would tell the truth. Fix the shape, and let the numbers move inside it. Write the page for the person deciding: a founder wants the verdict and the plan, a board wants the trend across quarters, and you want the learnings. One page serves all three when the verdict sits on top and the detail sits below. Resist writing for applause.
A good verdict line reads like a sentence you would say out loud: coverage was light this month, our citation rate for the two prompts we care about moved up, and here is the one change we are making. It states the mixed truth plainly and points forward. Keep it to a single sentence even when the month was complicated, because the detail below the fold is where the nuance belongs. Send the page on the same day each month and keep every past one. A single month is noise and twelve months is a trend, so the archive is where the real story lives. When you ask for budget, a stack of consistent reports makes a case no single dramatic chart can.
The hardest and most valuable habit is reporting an honest flat month. Coverage sometimes lands and little moves. A report that says so, names the likely reason, and proposes a change earns more trust than one that inflates a quiet month into a victory. Founders fund programs they believe, and they come to believe the person who told them the truth last quarter. State your confidence on each claim, and mark the signals that are strong, the ones that are noisy, and the ones you cannot yet read. A report built this way does more than record the past. It becomes the first half of the loop that makes next month's PR better.
Takeaway: Replace the clip dump with one honest page: a verdict, coverage quality, AI citations, demand signals, pipeline context, and next month's angles.
Chapter 6
Measurement that does not change next month's pitch is just bookkeeping. The point of all of this is a loop: measure the outcomes, learn what actually worked, and reinvest in more of it. Done well, each cycle sharpens the next one. Your list of targets gets better, your angles get stronger, and your coverage gets more relevant. Done poorly, or not at all, you keep pitching into the dark and calling the occasional hit a strategy.
Your own results are the best targeting data you will ever get. Score each contact on three things: did they reply, did they publish, and did their coverage move your signals or show up in an AI answer? A journalist who replies warmly but whose pieces never move anything ranks below one who rarely replies but lands coverage that gets cited. Let the record, not the relationship, set the order. Rank the proven ones up and send them more. The outlets that never landed, however prestigious their logos, move down. Over a few cycles your media list stops being a wishlist and becomes a ranked record of who engages with your story and whose coverage moves your signals. Keep the scoring light; three columns in a spreadsheet will do, and the discipline of filling them in beats any elaborate system you will not maintain.
Do the same with your message. Which angles pulled through into coverage in your own words, and which got rewritten into something you did not mean? Keep a short list of the buyer prompts you lose, the questions where an assistant names a competitor instead of you. Each lost prompt is a brief for a piece of coverage or a sharper proof point, and you win them one at a time. Write next month's angles from what the evidence says is landing, not from a blank page and a hunch. Share the winning angles with whoever writes your site and your pitches, so the message that earned coverage also shapes the pages an assistant reads when it decides whether to cite you. The loop turns every placement, and every miss, into direction for the one after it.
Reinvesting is a budget decision made with evidence. Put more of your hours into the outlets and angles that proved they move your signals, and stop spending them on the ones that only ever produced clips. For a lean team the scarce resource is attention, so the loop is really about pointing limited attention where last month's results said it pays off. The outlets you drop are not failures to mourn; they are attention returned to you to spend somewhere it works. Done every month, that one discipline compounds into a program that gets sharper while staying the same size.
Run by hand, this is a lot of tracking across forms, analytics, a media list, and a set of AI prompts, and lean teams let it slide exactly when they are busiest. That is the moment the loop usually breaks. Cites X is built to close it for you, honestly, with three connected parts:
The loop is only as good as its honesty. AI answers shift from week to week, vary with how a buyer phrases the question, and differ between people and tools. You cannot see every answer every buyer receives. Treat the citation reading as a compass that shows direction, not a scoreboard that promises a precise rank. The same humility applies to the whole loop: it points you toward better targets and better angles, it does not hand you certainty. That honesty is the feature. It keeps you investing in work that moves the business instead of work that moves a vanity number, which is the entire argument of this book in one habit.
Start small and start now. Pick the one business outcome you most need this quarter. Add the how did you hear about us field to your main form today. Set a branded-search baseline in Google Trends so you know what normal looks like. Write down the five buyer prompts you most want an assistant to answer with your name. Then take your next placement and track it end to end through the following month, across those four reads and those five prompts. That is the entire loop at small scale, and it will teach you more in thirty days than a year of clip counts ever did.
Takeaway: Measure, learn, reinvest, and start this week by tracking one outcome, one form field, one baseline, five prompts, and your next placement end to end.
Cites X runs the whole loop: build the brand, win the coverage, and become the answer AI gives.
Zugang anfragen → More Field Guides →