After this lesson you should be able to
- Build a coverage log that survives an audit months later
- Judge outlet quality without collapsing it to a single metric
- Calculate share of voice against a defined competitor set
- State clearly which outcomes cannot be attributed to a campaign
Decide what counts before you launch
The measurement argument in digital PR almost always happens after the campaign, when someone is disappointed. It should happen before, when everyone is still calm and nobody knows the result.
Agree in writing, before launch, which outcomes count as success: how much coverage, in what kind of outlets, with what proportion carrying links, over what period. Agree the competitor set you will compare against. Agree the reporting date, and a second one ninety days later, because coverage arrives late and links disappear. Most importantly, agree what will not be claimed — the effects that are real but cannot be attributed. That last conversation is uncomfortable exactly once, and it prevents a much worse one later.
The coverage log
Everything else is built on one artifact: a log with a row for every piece of coverage. Records recreated three months later are guesswork, so build it as the campaign runs.
| Field | Why it is there |
|---|---|
| Outlet and URL | The record itself; URLs change and articles move |
| Date published | Distinguishes the launch burst from the long tail |
| Original or syndicated | Separates independent decisions from republication |
| Linked or mention only | The primary split in the report |
| Link attribute and destination | Followed or nofollowed, and which page it points to |
| Prominence | Headline, body reference, or a passing name check |
| Journalist | Builds the contact list for the next campaign |
| Source | Pitched, syndicated, inbound, or follow-on |
| Status at 90 days | Still live, rewritten, link removed, page gone |
That last row is the one nobody builds and everybody needs. Links decay: on the profile I analyze throughout this course, the median referring domain stops linking after 1,080 days, and 97.4% of lost links disappeared while the source page was still perfectly reachable — the page stayed up and the link came out. A coverage report that is never re-checked describes a state of the world that no longer exists.
Judging outlet quality without one number
Resist the urge to reduce an outlet to a single authority score. Those metrics were built to describe link graphs, not journalism, and they rank a high-traffic aggregator above a specialist trade title that your entire buying audience reads. Score outlets on several dimensions instead and keep them visible separately:
- Audience fit. Do the readers include your customers, or people who influence them? A trade weekly can outrank a national paper on this.
- Editorial standard. Is there a masthead, named reporters, corrections policy, original reporting? Or is it a feed with a logo?
- Reach. Circulation or traffic, as an order of magnitude, not a false precision.
- Influence on other outlets. Does coverage here get followed? This is the dimension that produces second-order links and the one most reports ignore.
- Link characteristics. Followed or not, to which page, with what anchor. Last on the list deliberately.
Three or four tiers is enough. The point of the tiers is not precision; it is to stop a report in which forty automatic republications outrank the one placement that mattered.
Share of voice
Share of voice is your proportion of the coverage in a defined space over a defined period. It is the most useful campaign metric available because it is comparative — it accounts for the fact that some months are simply busier for your whole sector.
To calculate it honestly you need three decisions made in advance. Define the competitor set — named companies, not a vague category. Define the topic boundary — which stories count as your space. Define the source universe — which publications you are counting, and accept that it is a sample rather than everything ever published. Then count mentions of each brand in that space and express yours as a percentage.
Where it goes wrong: counting only the outlets you were mentioned in, changing the competitor set between reports, and counting a wire story reprinted forty times as forty mentions for you while counting a competitor's national feature as one. Weight syndicated copies down or count originals only, and apply the same rule to everyone.
Track the trend, not the number. Share of voice in a single month is noise. The direction over four quarters is a real signal about whether the program is changing your position in the conversation.
Links earned, mentions earned
Report these as two separate lines, never as one blended figure, because they behave differently and they mean different things.
Links earned should be counted by referring domain, not by link, and split into followed and nofollowed, original and syndicated. Counting individual links flatters a report enormously — on the profile I analyze in this course, the top 100 referring domains account for 84.3% of all links, and one domain alone sent 219,159 of them through a single sitewide placement. Any report that counts links rather than domains can be inflated by one lucky footer link into meaninglessness.
Mentions earned are the named references with no link. Historically these were treated as failures. That is no longer a sensible position: unlinked mentions in credible sources contribute to how search and AI answer systems associate a brand with a subject, which is the subject of Module 9. Count them, record where they appeared, and report them as their own outcome rather than as a consolation.
Also record where the links point. A campaign that earns fifty referring domains all pointing at a study page has built authority on a page with no commercial function. That is not a failure — internal linking distributes it — but it is a fact the report should state, because otherwise someone expects the money page to move and it does not.
What genuinely cannot be attributed
This is the section that separates an honest report from a persuasive one, and I would rather lose an argument than pretend otherwise.
- Ranking changes. You cannot isolate the ranking effect of a campaign. Rankings move for dozens of reasons at once, algorithms update continuously, competitors act, and there is no control group. You can report that rankings rose after a campaign. You cannot report that the campaign caused it, and a rank chart placed next to a launch date implies causation the data does not support.
- Advertising value equivalent. Pricing coverage as though you had bought the same space is an old PR habit and it is not defensible. Editorial coverage is not equivalent to advertising, the rate cards are notional, and the resulting figure exists to make a number look large.
- Brand search uplift. Branded search volume often rises after significant coverage, and that is worth reporting as a correlation. It is not clean attribution — other activity was running at the same time.
- Sales. Unless you have genuine last-click referral data from a specific article, attributing revenue to coverage is a story rather than a measurement. Report the referral sessions you can see and label the rest as unmeasured.
- Second-order coverage. A journalist who read the story and wrote about you six months later will not appear in any attribution model. It is one of the most valuable outcomes and it is effectively invisible.
- Presence in AI answers. You can observe whether your brand appears in generated answers, and it is worth tracking, but you cannot attribute that presence to a specific campaign.
Saying all this out loud makes the report weaker on paper and much stronger under scrutiny. Clients who have been oversold once recognize honest measurement immediately, and it is the reason they stay.
Questions
How do I prove digital PR improved rankings?
You cannot prove it, and claiming you can is where credibility is lost. There is no control group, algorithms change continuously, and competitors act at the same time. Report the links and mentions earned, report ranking and traffic movement alongside them as context, and describe the relationship as consistent rather than causal.
Should I use advertising value equivalent?
No. Pricing coverage as though you had bought equivalent advertising space is not defensible: editorial coverage and advertising are different things, rate cards are notional, and the number exists to look impressive. Report coverage, outlet quality, share of voice, links and mentions instead, all of which can be checked.
Do unlinked mentions count as a result?
Yes, and they should be reported as their own line rather than as a near-miss. Named references in credible sources contribute to how search systems and AI answer engines associate your brand with a subject. Some can also be converted to links later through reclamation, which is a separate and much cheaper piece of work.
When should I report on a campaign?
Twice. Once shortly after launch, when the initial burst has landed, and again at around ninety days to capture the long tail and, more importantly, to re-check what is still live. Links get removed while the page stays up — on the profile I analyze here, 97.4% of lost links vanished from pages that were still perfectly reachable.
Should I count links or referring domains?
Referring domains, in almost every case. Link counts are trivially inflated by sitewide placements: on the profile I use throughout this course, a single domain accounts for 16.8% of all links through one sitewide placement, and the top 100 domains account for 84.3% of the total. Domain counts describe how many independent sources chose to cite you.