In Muck Rack’s State of PR Measurement 2024, published on November 6, 2024, 40% of PR pros said they measure share of voice, and 20% said they do not trust it [Muck Rack, State of PR Measurement 2024].
You face that split each time you report the number. Share of voice gives you your mentions as a share of mentions for you and the rivals in a set and a window you pick.
You pick the rivals, keywords, platforms and dates. You change one pick and you change the result, with no change in the market. You count a spike in complaints and a spike in praise in the same way if you count volume on its own.
My view: you should read share of voice as movement inside one set that you hold in place, and split the count by channel and by sentiment before you act.
This post covers what share of voice counts, how you work it out, how a social count differs from a search count, and where you misread the number.
What share of voice counts#
I define share of voice as your mentions divided by mentions for you and your rivals, times 100. You count conversation in a place, on a topic, in a window. You get a percentage.
For you, the place is social and online conversation first. You count posts, comments, forum threads, news items, blog posts and review posts that name you or name a rival. You count each unit once under rules you set in advance.
Share of voice answers one question. Who takes conversation in this set, in this window?
You still need the count rules in writing:
You count a brand name, a handle and a product name that buyers use for you.
You count a post once, even if it names you twice.
You count a repost as a post if your tool counts reposts for rivals in the same way.
You drop spam, job posts and posts from your own accounts.
Write the rules. Change a rule and you restart the trend.
What share of voice does not count#
Share of voice counts volume in a set. It does not count sales, signups or revenue.
It does not count intent. A post that asks for a price and a post that mocks your price each add one to the count. You learn intent when you read posts and tag them.
It does not prove cause. Your share can rise in a week when a rival stops posting, when a rival has an outage, or when press covers your launch.
The percentage moves. The cause sits outside the percentage.
Hold that limit in mind and the number helps you. Drop the limit and you turn a count into a claim about the market. That move sits behind the distrust in the Muck Rack count in the intro [Muck Rack, State of PR Measurement 2024].
How you work out share of voice#
Follow the same order. Order protects the trend.
Step 1. Pick the set#
Pick three to five rivals that buyers name in deals and in search. Pick rivals you lose to, not brands you admire from a distance.
Lock the set for a quarter. Add a rival mid-quarter and you break the trend for that quarter.
Step 2. Write the keyword rules#
Write the name, handle and product terms for each brand. Add spellings buyers use. Drop terms that pull in a person, a place or a product with the same name.
Test each term on 20 posts. Read the 20. Drop a term if more than a few posts miss the topic.
Step 3. Pick the window and the places#
Pick a window that matches how you act. A week works if you post and reply each day. A month works if you report to a manager and plan content in months.
Pick places you can count in the same way for each brand: X, LinkedIn, Reddit, forums, news, blogs and review sites, as your access allows.
Note a gap in writing. A gap for one brand skews the share for all brands. Read what social media monitoring covers for place scope.
Step 4. Count, clean and divide#
Count mentions per brand. Read a sample from each brand. Cut spam, duplicates and posts from staff accounts.
Add the counts. Divide your count by the total. Multiply by 100.
Step 5. Split before you report#
Split the result by channel and by sentiment before you send it on. Report the total last, after the splits. I explain why in the split section below.
For example: a social count with numbers#
Take a 30 day window on X, LinkedIn and Reddit. You lock the set at you and three rivals. You use one keyword rule per brand.
Your counts after you clean spam and duplicates:
You: 300 mentions
Rival A: 400 mentions
Rival B: 200 mentions
Rival C: 100 mentions
Total: 300 + 400 + 200 + 100 = 1,000 mentions.
Your share of voice: 300 / 1,000 x 100 = 30%.
Rival A: 400 / 1,000 x 100 = 40%. Rival B: 20%. Rival C: 10%.
The four shares add to 100%.
Now read the next 30 days with the same set, rules and places:
You: 360 mentions
Rival A: 420 mentions
Rival B: 220 mentions
Rival C: 200 mentions
Total: 360 + 420 + 220 + 200 = 1,200 mentions.
Your share: 360 / 1,200 x 100 = 30%.
Your count rose by 60 mentions. Your share held at 30% because conversation across the set rose by 200. I would report both facts.
Volume rose, share held. A report that shows 30% in both months and hides the counts tells you half the story.
Change the set and the story breaks. Drop Rival C in month two and the total falls to 1,000.
Your share jumps to 36% with the same 360 mentions. You changed the denominator. You did not win conversation.
Callout: Worked example. 300 / 1,000 x 100 = 30%. You report the counts next to the percentage, so a reader can check the math.
How a social count differs from a search count#
A social count counts posts people publish. A search count counts queries people type.
In search, you count queries that name a brand. You divide your brand queries by brand queries for the set. This is share of search.
For example, in one month you log:
You: 8,000 brand queries
Rival A: 12,000 brand queries
Rival B: 6,000 brand queries
Rival C: 4,000 brand queries
Total: 8,000 + 12,000 + 6,000 + 4,000 = 30,000 queries.
Your share of search: 8,000 / 30,000 x 100 = 26.7%.
Share of search represented 83% of share of market on average across 30 cases in 12 categories and 7 countries [IPA Share of Search Think Tank, 2021].
You read that link in search data, in those cases. A social count earns no such link on its own.
Social posts include complaints, support asks, jokes and bots. Search queries show a person looked for a brand by name.
Use search as a contrast. If your social share rises and your search share holds, conversation rose around you and demand held in place.
If search rises after social, posts may lead queries. I start with two or three cycles before I claim a pattern, and I adjust that count for my volume.
A 10 point gap between share of voice and share of market linked to 0.5 points of market share growth in a year, on average, across 123 brands in 30 categories [Nielsen, 2009].
Use it as a planning reference for paid voice. Do not apply 0.5 to a social mention count.
Split the count by channel and by sentiment#
I put more weight on splits than on the total. The total hides the move you need to act on.
Split by channel first:
X moves in hours. A reply, a quote post or an outage can swing a day.
LinkedIn moves in work hours. Posts from staff and buyers carry weight in B2B sets.
Reddit moves in threads. One thread with 80 comments can lift a week.
News and blogs move in days. One item can drive posts in other places.
Review sites move in months. A run of posts after a price change shows up here first.
Return to the worked example. Your 300 mentions split as 120 praise, 60 complaints, 90 questions and 30 neutral posts.
Rival A has 400 mentions with 200 complaints after an outage. Rival A leads on volume. You lead on praise.
Act on volume alone and you copy a rival in a crisis week.
Report three lines each cycle:
Share by channel, with counts.
Share by sentiment, with counts.
Share in total, with counts.
A reader can check your math from counts. A percentage with no counts asks a reader to trust you.
How you pick rivals, keywords and a window#
Pick rivals from buyer language. Pull names from sales calls, support tickets, review sites and posts that ask for an alternative. If buyers name a brand when they drop you, that brand belongs in the set.
Cap the set at five. Past five, each share gets thin and a small count swings the result.
Park extra brands in a watch list. Promote one in the next quarter if buyers keep naming it.
Write keywords as buyers write them. Add a handle, a name with no space and a spelling that shows up in posts.
Test a term that matches a first name or a city name with a second term, such as a product word. Read posts until the misses drop.
Start with a floor of 200 mentions in your window, and adjust that floor for your volume. Under 200, one thread moves the share by points.
In a category that hits your floor in 90 days, run 90 days and read at 90 days. In a category that hits your floor in 7 days, run 7 days for action and 30 days for reporting.
Lock the picks. Log each change with a date and restart charts from the change date.
I show a trend from the change date. I drop a chart that runs across a change with no date note.
If you track conversation in a tool, keep the same set, terms and window in the tool. You can track in CommunityTracker in that way, with the same rules you wrote for a manual count.
Tool choice does not fix set choice. Rules fix the read.
For a wider view of tool scope, see social listening tools.
How often you read it, and how you act#
I act after movement holds for 2 cycles in a row in a split, and I treat 2 cycles as a starting point I adjust for my volume. I skip a one day spike in the total.
Callout: Starting points. Start with 200 mentions in your window and 2 cycles in a row, and adjust both for your volume.
Use this order when share moves:
Check the denominator. Did the set total rise or fall, and which brand drove it?
Check the channel. Which place drove the move?
Check sentiment. Did praise move, or did complaints move?
Read 20 posts behind the move. Name the topic in one line.
Pick one act: reply, post, brief a writer, brief sales, or watch one more cycle.
For example, your LinkedIn share rises from 22% to 31% in a month, and questions drive the rise. Read the questions.
If buyers ask how you handle a use case, write a post that answers that use case and give sales the link. Count again in the next cycle. Keep the act tied to the topic that moved.
If your share falls and a rival share rises on praise for a feature you lack, do not chase the feature in a week. Log the posts, count buyers who name the feature and take that count to product planning.
Share tells you where buyers talk. It does not set your roadmap on its own.
I disagree with a habit I see in teams here. Teams set a target such as 40% share and push posts until the count hits it. That target rewards volume.
It rewards replies that add nothing and posts that start rows. Set a target on a split you value, such as share of questions on LinkedIn or share of praise on review sites, and pair it with a count floor so a set under your floor cannot pass as a win.
Mistakes that break the read#
Volume with no sentiment check tops the list. A crisis can double your mentions in a day. Your share rises and your position worsens.
Bots and accounts that post on a loop can inflate a count, and a count that rises on complaints harms a brand [Muck Rack, Share of Voice: Common Mistakes, 2024].
Watch for these breaks:
You change rivals, terms or places and compare across the change.
You count reposts for you and drop reposts for a rival.
You count your staff posts, launch posts from partners and posts in a giveaway with no tag.
You report a percentage with no counts, no window and no set list.
You average channels and hide a fall on the channel buyers use most.
You chase a spike that sits in one thread, one day or one account.
You treat share of search, share of posts and share of spend as one number.
Bot and spam posts need a rule in writing. Cut accounts that post the same link in a loop, posts with no text beyond a link and a handle, and duplicates that a tool copies across places.
Sample 50 cuts. If you cut buyer posts by error, fix the rule and recount.
Write a note next to each spike. Write the event, the date and the channel next to the chart. A launch, an outage, a price rise and a press item each explain spikes in a way a trend line cannot.
Who should track share of voice, and who should skip it#
Track share if buyers talk in public and name rivals when they talk. B2B software, consumer products with review sites and categories with active forums give you posts to count. You need a rival set buyers repeat and a cycle where you can act.
Skip share as a headline number if buyers buy in private, sign through partners or talk in closed groups you cannot count. Count what you can audit, such as inbound asks that name a rival, win and loss notes and review site posts. Report those counts in plain form.
Start with a floor of 200 mentions in 90 days, and adjust that floor for your volume. Skip share if your set sits under your floor in 90 days.
The math works and the read fails. One thread will run your quarter.
Use queries, interviews and review reads until volume grows.
Skip a paid voice target if you have no market share base to compare it with. Excess share of voice needs market share in the math. With no base, track social and search share as trend lines and leave growth claims out.
FAQ#
What counts as a mention?#
A mention is a post, comment, thread, news item, blog post or review post that names you or a rival under your keyword rules. You count each unit once and you drop spam and staff posts.
How many rivals should you include?#
Include three to five rivals that buyers name. Add a brand when buyers name it in deals, posts or reviews in cycle after cycle.
Should you aim for 100% share of voice?#
No. A set with rivals will split conversation. Aim for movement in a split you value, such as questions or praise in a channel, and hold the set and rules in place.
Final thoughts#
Share of voice gives you a share of conversation in a set you define. That limit gives the number value.
I read share as movement, in splits, with counts on show. I hold the set, terms and window in place. I act after movement holds for 2 cycles in a row and after I read the posts behind it, and I treat 2 cycles as a starting point I adjust for my volume.
Start this week with three rivals. Write the terms, count 30 days and split by channel and by sentiment. Report counts next to each percentage.
You will learn more from that one count than from a year of percentages with no counts.