Making press coverage a measurable signal

A PR campaign ends with a coverage report: a list of links, often with circulation figures attached. Everyone reads it once, and it never appears in the same place as any other number about the project.

Which means when a week moves, press is not among the explanations considered, and the movement gets credited to something else.

Turn it into a series

Press becomes analytically useful the moment it is a dated count rather than a list.

Per piece: date published, outlet, headline, link. That is enough to produce a count per day or per week, and a count per week put next to streams, search demand and ticket sales is where the value is.

Coverage clusters, because a campaign lands in a fortnight and because outlets follow each other. Knowing that five pieces published in the last seven days is what turns "an odd week" into "the week the campaign landed".

What press should move, and how you would know

Branded search demand. The best available proxy for reach beyond your existing audience. Somebody who read a review and typed the artist's name into a search engine is someone press reached. Google Search Console shows the impressions and clicks for your own site, and a step in branded queries within a couple of days of a piece is about as close to a measurable press effect as this business gets.

Landing page arrivals from that outlet. If you gave the outlet a tagged link, this is direct. Most PR does not do this and it costs nothing to start.

Streams and follows. Slower, smaller and harder to isolate, because press rarely runs alone.

Ticket sales in the outlet's region. For local press around a show, this is the one that matters, and the timing is tight enough to be readable.

The automated monitoring problem

Monitoring coverage automatically, for example from a news feed, is cheap and has one predictable failure: common names.

A project called something generic will pull in unrelated articles about a phrase rather than an artist. So two things are needed.

A per-project search term that can be overridden. Defaulting to the project name is right most of the time and wrong badly enough often enough that an override has to exist. Ours falls back to the project name and lets an admin set a specific query where the name is generic.

Human judgement before citing. Any automated feed will occasionally surface a same-name article, so anything that reads the feed, including an AI assistant, has to be told to check the title, outlet and date rather than assuming every result is about the project.

What not to do

Do not compute a sentiment score at low volume. Six articles a month cannot support a sentiment metric that means anything, and a number moving from 0.62 to 0.58 will get discussed as though it did.

Do not use circulation or reach figures as an outcome. They are the outlet's audience, not yours, and they are the press equivalent of impressions.

Do not judge PR on volume alone. Ten aggregator reposts of one story is one story. A single well-placed piece in a market where you have a show is worth more than a national round-up, and only the dated, outlet-attached version of the data lets you tell those apart.

The practical setup

Automated monitoring for the count and the timeline, plus your PR company's own coverage report for accuracy and for anything the feed missed. The two together are better than either: the feed gives you dates and consistency, the report gives you judgement and the pieces that never appeared in a news index.

Then put the count on the same chart as everything else, and stop treating press as a separate document that arrives at the end.