What is share of voice?

The calculation, plainly

The arithmetic is a fraction:

Share of voice = (your mentions / total mentions across the whole competitor set) x 100

The numbers below are illustrative arithmetic, not measurements of anything. Suppose that across one month, on the sources you chose, you count 120 mentions of your brand, 300 of competitor A, 60 of competitor B and 20 of competitor C. The total is 500, so your share of voice is 120 divided by 500, which is 24 percent.

That is the whole calculation, and the reason it still confuses people is that the fraction hides three arbitrary decisions: who is in the denominator, which sources are counted, and how long the window is. Change any one and the answer changes without anything happening in the market.

The variants, and where the data comes from

Share of voice is a shape rather than a single metric. The version you want depends on the question.

Variant	What it counts	Where the data comes from	The trap
Mention share	Brand mentions across monitored sources	A media monitoring or social listening tool	Two firms with similar names, and mentions of a common word
Search share	Impressions your pages earn against the total available for a keyword set	Search Console impressions, plus keyword volume data	The keyword set you picked defines the answer
Paid share	Impression share on your ad keywords	The advertising platform reports it directly	It measures budget as much as it measures demand
Review share	Your review count against the competitor set's	Manual counting from the listings	Count is not quality, and old reviews still count
Answer share	How often you are named in AI assistant answers to category prompts	Manual prompting, logged over time	Answers vary between sessions, so one run is a sample

Answer share is the newest of these and the least standardised. It is worth tracking because it is where a growing share of category questions are being asked, and because a name absent from those answers is invisible in a way no rank tracker will show you.

What determines the number, before the market does

Three choices set the result, and they need to be recorded next to the metric or the trend is meaningless.

1. The competitor set. A set of three large rivals produces a small share for you. A set of thirty regional operators produces a larger one. Neither is dishonest, and neither is comparable to the other.
2. The source set. Adding news to a social only measure changes the balance toward whoever gets press. Adding forums changes it toward whoever people actually discuss.
3. The window. Short windows are volatile and useful for campaigns; long windows are stable and useful for strategy.

Freeze all three, then report the number monthly. Compare it only to your own prior numbers on the identical definition. A share of voice figure taken from someone else's report, computed on their competitor set with their tool, is not something you can benchmark against, and vendors quoting industry average share of voice are quoting the output of their own configuration.

What share of voice tells you

Used carefully, it earns its place in a small set of situations.

• Category entry. Starting near zero and watching the fraction move is a legitimate measure of whether anyone has noticed you exist.
• Competitive gaps. The breakdown matters more than your own number. Seeing that one rival owns most of the forum conversation while another owns the news tells you where the ground is.
• Campaign effect. A defined baseline before a campaign and the same measure after is one of the few clean readings available in reputation work.
• Topic ownership. Share of voice on a specific subject, rather than on your brand name, shows whether you are associated with the thing you want to be known for.

What it does not tell you

This is the part that gets left out, and it is where the metric does its damage.

It says nothing about whether the talk is favourable. Volume is not approval, and the largest share of voice a company ever records is usually during its worst week. That is the whole reason brand sentiment is reported alongside it rather than instead of it, and reporting either alone produces a misleading picture.

It says nothing about the audience. Ten thousand mentions in a community that will never buy from you is a worse result than two hundred mentions among people who will. It says nothing about whether the mentions refer to you at all: entity disambiguation, deciding whether a name in a sentence refers to your company, is a genuinely hard natural language problem with an extensive research record indexed in the ACL Anthology of natural language processing research, and monitoring tools solve it imperfectly.

It also has no bearing on whether anything said about you is actionable. A spike driven by a specific false statement of fact is a different question entirely, governed by defamation law rather than by measurement, and Cornell Law's entry on libel is a better starting point than any dashboard for understanding what that involves.

Single number metrics attract the same criticism wherever they turn up, and Nielsen Norman Group's critique of Net Promoter Score is the clearest short statement of the general problem: a figure compact enough to travel through an organisation travels without the caveats that made it meaningful.

And it is not a target. Share of voice is trivially inflatable with paid placement and volume publishing, which is why it is a poor incentive to attach to anyone's compensation. Measured honestly it is a thermometer; managed directly it becomes an exercise in making the thermometer read what you want.

Using it without misleading yourself

Report it with four things attached: the competitor set, the source set, the window, and the sentiment split. Chart it against your own history rather than against a benchmark you cannot audit. Treat a large single month move as a question rather than a result, and go and read the mentions that caused it, because the reason is usually one thread rather than a trend.

If the number is moving and you cannot see why, the answer is in the underlying material rather than the chart. Reading that material, the results, the reviews and the sources behind them, is what a reputation audit does first.