Marketing & GrowthFree
Share of Voice Analysis With Deduped Clips
De-duplicate the copies, admit the denominator is a choice, then count how often your own messages survive into the coverage.
Halveston's monitoring tool reported 412 mentions in the quarter and a 19.4% share of voice, up three points. Collapsing syndicated copies of the same originating report left 148 distinct reports and a syndication ratio of 2.78. Across the whole category the exercise removed 889 of 2,121 mentions, 41.9%. Halveston's share of voice fell to 12.0%, because it ran two wire releases in the quarter and its largest competitor ran none at all. The gain was a distribution contract.
Then the denominator, which is a choice rather than a fact. The same 148 reports give a 17.5% share against the four named vendors, 12.0% against every vendor named in the category, and 7.6% against the whole category term. That is a ten point spread, 3.3 times the three point gain being reported as progress. The number is quoted as a range and the query behind each bound is written down. A single figure with no query attached is not a measurement.
Written for whoever has to answer whether PR is working with something other than a clip count. Pair it with the press release kit when the next announcement is what you are measuring, and the competitor teardown when a rival's message keeps landing and you want to know why. The ad and messaging teardown reads the paid surface the same way, and the page copy rewrite is where a message that never lands gets rebuilt.
What a clip count is counting
A wire release becomes many URLs, and a monitoring tool counts URLs. Google's own guidance on article duplication says that where there are many versions of the same article, it can be difficult for our systems to identify the original article. If that is hard for the company with the largest news index, a monitoring dashboard is not quietly solving it. Collapsing to distinct originating reports is manual work, and here it removed two fifths of the raw count.
Every share metric is defined by its denominator, and the good ones say so. Google's documentation for Trends states that each data point is divided by the total searches of the geography and time range it represents before anything is compared. A media share of voice skips that disclosure entirely. Nobody records what the category total was or which boolean query produced it, which is why the same quarter supports three different answers.
The statistic that needs no denominator is whether your messages appear. Of Halveston's 148 reports, 54 carried at least one of its five messages, 36.5%, and 94 carried only the company description. Its best-performing message reached 31 reports, but 24 of those came from one analyst briefing, so without that single event it sits in 7 of 148. Wexcombe had the least coverage in the set and the highest penetration at 61.2%. It ranks last on clips and second on message.
How it works
Collapse the copies
Group syndicated versions of one report together, then count originating reports rather than URLs.
Bracket the share
Three defensible denominators, three answers, and the boolean query recorded beside each one.
Count the messages
Each named message marked present or absent in every report, yours and theirs.
Rank on penetration
The order by message-carrying reports, which is the only one that tracks positioning.
What you get
- Raw mentions collapsed to distinct originating reports, with the syndication ratio per vendor.
- Share of voice as a range across three denominators, each with its query.
- Every one of your messages counted across your own coverage, report by report.
- The same message count for each competitor, on their own coverage, for comparison.
- Reports carrying nothing but the company description, which is usually the majority.
- A Sheet of every report, a Doc of the reading, and the trend chart.
Common questions
Isn't share of voice just mentions divided by category mentions?
That is the formula, and the problem is that both terms are unstable. The numerator counts syndicated copies, which inflated Halveston's by 2.78 times. The denominator is whatever a boolean query returned, and three defensible ones gave 17.5%, 12.0% and 7.6% for the same quarter. A ten point spread is not a metric.
So should we stop reporting share of voice?
Report it as a range with the query behind each bound, and stop reporting quarter-on-quarter movement smaller than the range. Halveston's three point gain sat inside a ten point spread. Then put the message-carrying count next to it, because that one is a rate within your own coverage and needs no category total at all.
Does de-duplication punish us for using a wire?
It stops a distribution contract from reading as a change in coverage. The wire is doing its job, and the syndication ratio is reported openly beside the count so the effect is visible rather than removed. Halveston's ratio was 2.78 against 1.59 for its largest competitor, which is most of the reported gain.
What counts as a message appearing?
The idea reaching the reader, in the reporter's words or in a quote, not the phrase matching exactly. A release saying month-end close drops from days to hours and an article saying the same thing differently both count. A company description, a funding figure and a headcount do not, which is what 94 of the 148 reports carried.
Why score competitors on their own messages?
Because it is the only fair comparison and it changes the order. Corbridge carried a message in 53.0% of its reports and Wexcombe in 61.2%, against Halveston's 36.5%. Wexcombe has the least coverage in the set and the most message penetration, so the smaller programme is the one working.
How many messages should we be tracking?
Between four and six, written verbatim as your positioning states them. More than that and penetration is diluted across ideas nobody is actually pitching. Halveston tracked five, and two of them appeared in 6 and 3 reports out of 148, which is a finding about the pitch rather than about the press.
What actually gets delivered?
A Sheet with one row per report, carrying the outlet, the tier, the originating report it belongs to, the vendor and which messages appeared. A Doc with the bracketed share, the penetration table and what moved. A chart of both series against the competitor set. The next announcement belongs in the press release kit.
Share of Voice Analysis With Deduped Clips
Fill in the form and your workspace opens with the work already underway.