Analytics · September 28, 2026 · 8 min read
Share of Voice: Measuring Category Visibility Instead of Rankings
A visibility number that falls when a competitor wins beats any average position report — here is the keyword universe, the weighting, and the reports behind it.
By FluxWriter
Share of voice is the metric that tells you whether you are winning a category, and average position is the metric that tells you almost nothing. Most reporting still leans on 40 hand-picked keywords and a monthly average nobody can explain afterwards. This covers what the number actually measures, the weighting models worth using, the keyword universe that makes it defensible, and the reports that produce it.
Why Average Position Flatters a Losing Quarter
Average position hides the two things that decide traffic — how much each position is worth, and which queries you are counting. A page moving from 3 to 2 earns roughly 40 to 50% more clicks. A page moving from 18 to 12 earns nothing extra, and the average records the second move as the bigger win.
Every impression carries equal weight in that average, so a query you rank 40th for on one obscure search counts as much as your best commercial term. Add 20 long-tail queries where you sit at position 8 and the average improves. Traffic does not.
The second failure is scope. A tracked list of 40 keywords is a list you chose — and you chose it back when the category looked different. Competitors launch pages against terms you never added, and your average holds steady at 7.4 while three rivals take positions 1 through 3 on the 60 queries you were not watching.
That is the gap this metric closes. It asks a different question. Of all the clicks available in this category this month, how many landed on you? A number that drops when a rival wins tells you more than one that only moves when you do.
What Share of Voice Actually Measures
Share of voice is your slice of the estimated clicks available across a defined set of queries, stated as a percentage. It needs a query set, a search volume for each query, and a click-through curve that turns a ranking position into an estimated click count.
The arithmetic is unremarkable. Estimate the clicks your positions earn across the set, estimate the clicks the whole set generates, divide. If the clicks available across your category add up to 400,000 a month and your positions imply 18,000 of them, your share sits near 4.5%.
That percentage matters less than its direction. A 4.5% share is neither good nor bad on its own — it is a baseline, and the value arrives in month 3 or 4 when you can see whether it climbs while the category itself grows.
Two properties earn the setup cost. The number is bounded, so any gain has to come out of somebody else, which forces the competitive question into every report you write. It also compares across categories in a way keyword counts cannot — 200 top-ten rankings against 50 serious competitors means something very different from 200 against 4.
Choosing a Weighting Model
The weighting model decides what your number rewards. Pick the wrong one and the metric improves in months when revenue does not. Score the four options against what you are trying to prove:
| Model | What it counts | Best for |
|---|---|---|
| Unweighted position share | Rankings held, every query equal | Small sets under 100 terms |
| Volume-weighted | Search volume behind each position | Category-level trend reporting |
| Volume × CTR curve | Estimated clicks, position-adjusted | The default for most teams |
| Click-value weighted | Estimated clicks × cost per click | Commercial and ecommerce sets |
The step from row one to row two is the one that changes conclusions. Unweighted share treats a 90-search-a-month query and a 9,000-search one identically, which is exactly how a report ends up celebrating movement on terms nobody types.
Row three is where most teams should land. It costs you one extra column in the spreadsheet and it stops position 6 and position 2 from being scored as near neighbours, when the second earns roughly 3 to 4 times the clicks of the first.
Leave row four alone unless someone above you wants a currency figure. Click-value weighting holds up in categories where cost per click genuinely ranges from $3 to $40 and that spread reflects real buying intent. Where CPC is flat across the set, it adds noise and an argument you will have to defend for no gain.
Defining a Keyword Universe You Can Defend
The universe is the whole exercise. Get it wrong and every downstream number is decoration.
Three rules keep it honest. The category defines the set, not your rankings — the terms you lose on have to be in there. A universe assembled from your own Search Console queries guarantees a flattering result, because it contains only the searches you already surface for.
Freeze it. A set that quietly grows every month cannot be compared against itself, and a share that drops 2 points because you added 80 hard terms is a reporting artifact, not a loss. Add terms once a quarter as a dated batch, then recalculate the previous 3 months against the new set so the trend line survives.
Size it deliberately. Somewhere between 300 and 1,500 queries covers most single-category sites, and a first build usually lands nearer 300 to 800. Under 200 terms, one volatile head keyword swings the entire percentage. Above 3,000, you are paying tracking fees on terms with no bearing on the outcome.
Build the set from the queries your top 5 competitors rank in the top 10 for, then add the search terms report from your paid campaigns if you run any. Sales calls fill the gap the tools leave — the questions that keep coming up rarely show volume, and they convert anyway. Strip branded terms out before you calculate anything.
Where the Number Comes From
No tool hands you an accurate share of voice for free, and every source trades something away. Name the report before you quote the figure.
Ahrefs Rank Tracker reports share of voice directly against a tracked list and draws competitor lines on the same chart. Semrush Position Tracking reports a Visibility percentage across your set, with estimated traffic beside it. Both are modelled from positions. Neither has seen a click.
Search Console has seen every click you earned, and none of anyone else's. That makes it the calibration layer rather than the source. Pull the Performance report over the full 16 months it retains, set modelled clicks beside actual clicks for the same queries, and revisit the curve if the two sit more than 25 to 30% apart.
The vendor's published click-through curve is good enough for month 1. By month 6, a curve fitted to your own Search Console data will be closer — building one is a separate exercise, but the rule that matters here is to strip branded queries out before you fit anything. Left in, they can push position 1 above a 40% click-through rate and distort every position under it.
What the Number Cannot Tell You
Share of voice is an estimate stacked on two other estimates, and presenting it as measurement is the fastest way to lose a room full of skeptics.
Search volume is modelled, not counted. Two vendors will often disagree by a third or more on the same term, so an absolute share only means something inside one tool. Comparing a percentage from Ahrefs against one from Semrush proves nothing — the trend within a single tool is the deliverable.
Positions are sampled from one location and one device profile. If your category splits across mobile and desktop, or across 12 metros with different local results, a single tracked profile describes none of them accurately.
The curve also assumes clicks that increasingly do not happen. AI Overviews, featured snippets and map packs absorb clicks the curve still credits to position 1. Where an answer panel sits on most head terms, modelled clicks commonly overstate reality by a third or more. Report the percentage next to real Search Console clicks, never in place of them.
FAQ
How often should I recalculate this?
Monthly for reporting, quarterly for decisions. Weekly readings mostly capture ranking noise — a 500-query set will typically swing 1 to 2 points week to week with nothing real behind it. Set the reporting cadence at 4 weeks and the universe review at 13 weeks, and date both in the report.
Can I use share of voice on a brand-new site?
No, and forcing it early wastes a quarter of reporting. Below roughly 2%, the figure is dominated by rounding and one lucky ranking. Track top-10 counts and impression growth instead until your share reads above 3 to 5% for 3 straight months, then switch the report over.
Should branded queries stay in the set?
No. Branded terms sit at position 1 with click-through rates often above 40%, so including them inflates your share by several points and buries every competitive movement underneath. Keep a separate branded line on the same dashboard if you want it, never inside the category universe.
The Practical Takeaway
Build the universe before you build the metric. Assemble 300 to 800 category queries from what your top 5 competitors rank in the top 10 for, strip out branded terms, freeze the set under a dated version, and pick volume-times-CTR weighting unless a finance team is asking for a currency figure. Calculate it monthly, review the universe every 13 weeks, and print the percentage beside your actual Search Console clicks so the estimate never travels alone. Start with one category — the one your revenue depends on — and leave the rest untracked until that first set has 3 months of trend behind it.
If you are publishing against a category universe that wide, tools like FluxWriter can help keep coverage moving on the terms you have already decided are worth owning — but choosing which queries define the category, and defending the weighting behind the number, stays a judgement you have to make yourself.