Share of voice: the measurement system for smarter budget shifts

A quarterly share of voice chart can become useless with one quiet spreadsheet edit.

The marketing team starts the quarter tracking five competitors. A new rival appears in week seven, so someone adds it to the dashboard. Reported share of voice falls from 18% to 14%. Leadership reads the decline as lost visibility. In reality, the denominator changed. The trend no longer compares like with like.

This is a common operating failure because teams treat share of voice as a single percentage. It is a measurement system. Search impressions, social mentions, press coverage, and AI recommendations each answer a different management question. Their numerators differ, their source data differ, and their commercial meaning differs.

The discipline starts with two rules:

Freeze the competitor set, query set, source set, geography, and time window for the reporting period.
Model commercial impact from your own conversion history rather than attaching a universal revenue claim to an SOV increase.

A move from 10% to 15% share of voice is an increase of 5 percentage points and a 50% relative increase in measured visibility. Both figures are correct. Neither tells you the pipeline effect without evidence from your funnel.

This guide explains how to calculate channel-level SOV, build a reporting workflow, set useful benchmarks, add AI recommendation visibility, and use the result to make budget decisions.

What is share of voice?

Share of voice is your brand's measured visibility divided by the measured visibility of all brands in a defined competitive set.

The general share of voice formula is:

SOV (%) = your brand visibility ÷ total visibility for the tracked brand set × 100

The word “visibility” needs an exact definition. Depending on the channel, it may mean impressions, mentions, estimated organic clicks, press articles, citations, or recommendations in AI answers.

For example, if your brand receives 2,400 mentions and the six tracked brands receive 12,000 mentions in total, your social mention SOV is:

2,400 ÷ 12,000 × 100 = 20%

That answer means your brand produced one-fifth of measured social mentions within that fixed dataset. It does not mean you have 20% market share, 20% of social engagement, or 20% of revenue.

Four principles keep the metric useful:

Define the numerator in channel-native terms.
Use the same competitor set and measurement window within each channel.
Report absolute and relative change together.
Pair visibility with a downstream measure such as qualified traffic, opportunities, win rate, or revenue.

A useful share of voice definition therefore includes its boundaries. “We have 17% SOV” is incomplete. “We have 17% organic search visibility across 420 non-branded commercial keywords in the UK, measured against five frozen competitors during Q1” can support a decision.

Share of voice, share of market, and share of search are different measures

These terms are often placed next to each other as if they were interchangeable. They are not.

Share of voice measures your portion of visibility within a specified dataset. It is usually an input or intermediate indicator.
Share of market measures your portion of category sales, units, customers, or revenue. It is a business outcome.
Share of search measures your brand's portion of branded search demand within a category. It can act as a signal of current interest or mental availability, subject to the category and data quality.
Impression share usually means impressions received divided by impressions for which the ad or listing was eligible. This is common in paid search and does not necessarily use competitor impressions as its denominator.
AI recommendation share measures how often an AI system recommends your brand relative to tracked alternatives for a controlled prompt set.

Byron Sharp's How Brands Grow provides useful context for mental and physical availability. That context helps explain why being easy to recall and easy to buy matters. It does not turn every visibility metric into a sales forecast.

The IPA publication The Long and the Short of It is a useful source for the planning concept behind excess share of voice, meaning share of voice relative to share of market. Apply that idea as a planning lens, then test it against your category economics. Do not convert it into a universal promise that a given SOV increase will produce a fixed market-share gain.

Why one blended SOV percentage leads to poor budget choices

A blended score can conceal the exact gap the team needs to fix.

Suppose a company reports 22% “total SOV.” Organic search is strong, paid impression share is falling, social mentions are inflated by a giveaway, press coverage is flat, and AI systems rarely recommend the product for high-intent category questions. Averaging those figures creates a tidy number with little management value.

Each channel answers a separate question:

Search SOV asks how visible the brand is for a defined keyword universe.
Paid impression share asks how often ads appeared when eligible, with budget and rank losses explaining missed exposure.
Social SOV asks how much tracked conversation refers to the brand.
PR SOV asks how much relevant editorial coverage includes the brand.
AI recommendation SOV asks how frequently the brand appears as a suitable answer to controlled user questions.

These measures can sit on one dashboard, but they should remain separate. If leadership wants a single executive index, normalize each channel score, state the weights, and preserve the underlying channel figures. Never add raw impressions, mentions, articles, and AI answers into one denominator. Those units have no common mathematical meaning.

This distinction changes budget allocation. A search visibility gap caused by weak category pages needs content and technical work. A paid impression gap caused by budget limits may need more spend. An AI recommendation gap caused by weak third-party evidence may need citation building, clearer product facts, and off-site reputation work. Sending the same extra budget to each channel would be administratively simple and commercially careless.

Channel-specific share of voice formulas

Organic search and SEO share of voice

SEO share of voice should use a fixed set of keywords, locations, devices, and competitors.

A common visibility-based formula is:

SEO SOV = your weighted organic visibility score ÷ total weighted visibility score for all tracked brands × 100

A weighted visibility score can assign more value to high-volume keywords and higher positions. The exact click-through assumptions must remain stable for trend reporting.

You can also use estimated organic clicks:

Organic click SOV = your estimated clicks from tracked keywords ÷ total estimated clicks for the tracked brand set × 100

Worked example:

Your estimated clicks from the fixed keyword set: 18,000
Competitor A: 27,000
Competitor B: 21,000
Competitor C: 14,000
Total: 80,000
Your organic click SOV: 18,000 ÷ 80,000 × 100 = 22.5%

Use this for relative search visibility. Use analytics and revenue attribution for actual commercial results. Estimated clicks from an SEO tool are modelled values, not observed sessions.

Paid search share of voice

Paid media teams commonly use impression share:

Paid impression share = impressions received ÷ total eligible impressions × 100

If an account received 420,000 impressions and was eligible for an estimated 600,000, paid impression share was 70%.

This metric answers an availability question: how often did the ad appear when it could have appeared? Lost impression share due to budget and lost impression share due to rank then help explain the missing 30%.

If competitor impression data comes from an auction-insights source, keep that analysis separate from platform impression share. One uses eligibility as the denominator. The other compares auction overlap or relative presence. Calling both “paid SOV” without labels invites bad decisions.

Social share of voice

The basic formula is:

Social mention SOV = your qualified brand mentions ÷ qualified mentions of all tracked brands × 100

Assume the brand has 3,100 qualified mentions and the peer set has 15,500. Social SOV is 20%.

“Qualified” needs a written rule. Exclude spam, bot-like repetition, recruitment posts, employee reposts, irrelevant uses of an ambiguous brand name, and campaign mechanics that distort ordinary conversation. Report engagement separately. A brand can have high mention SOV and weak audience response.

You may also calculate engagement SOV:

Social engagement SOV = engagement on your tracked brand content ÷ engagement on all tracked brand content × 100

Mention share and engagement share answer different questions, so keep both labels visible.

PR and earned-media share of voice

A basic PR formula is:

PR coverage SOV = qualified articles mentioning your brand ÷ qualified articles mentioning all tracked brands × 100

If the brand appears in 34 qualified articles and the peer set appears in 170, PR coverage SOV is 20%.

Article count alone can reward low-value syndication. Add separate quality fields such as publication relevance, article prominence, sentiment, named spokesperson inclusion, and link or citation presence. If you create a weighted PR score, publish the scoring rules internally and do not change them mid-quarter.

PR reach is another possible numerator:

PR reach SOV = estimated reach of your qualified coverage ÷ estimated reach of all qualified peer coverage × 100

Treat estimated reach carefully. Publisher audience estimates and actual article readership are different quantities. Use the estimate for directional comparison, not as observed exposure.

AI recommendation share of voice

AI visibility requires a controlled query set, a named model set, a fixed location or language where relevant, and repeat runs because answers can vary.

A straightforward formula is:

AI recommendation SOV = your brand recommendation occurrences ÷ all recommendation occurrences for tracked brands × 100

Suppose 100 controlled AI answers produce 160 total recommendation occurrences across your peer set. Your brand appears as a recommendation 24 times.

24 ÷ 160 × 100 = 15% AI recommendation SOV

Track at least two related measures:

Recommendation rate: the percentage of eligible answers that recommend your brand.
Citation share: your portion of cited sources or linked domains associated with the category.

A mention may be neutral, negative, or incidental. A recommendation says the system presented the brand as suitable for a stated need. A citation says the system used or referred to a source associated with the brand. These events belong in separate fields.

For teams building this measurement, AI visibility tracking can monitor brand presence and competitor presence across controlled queries. Citation tracking then helps distinguish being named from being used as a source.

Share of search

Share of search usually uses branded query volume:

Share of search = searches for your brand ÷ searches for all category brands × 100

Keep generic category searches outside the numerator unless your method explicitly defines a different measure. Correct for brand-name ambiguity, spelling variants, product names, and major events that create temporary search interest.

Share of search can complement SOV reporting. It should not replace channel visibility or conversion data.

How to quantify change without overstating it

Percentage-point change and relative change should appear together.

If SOV rises from 10% to 15%:

Absolute change: 5 percentage points
Relative change: (15% - 10%) ÷ 10% = 50%

Calling this a “50-point increase” would be wrong. Calling it “up 5%” would be ambiguous. Write both numbers.

Commercial impact needs a separate model. Use the company's observed path from exposure to revenue, with ranges where causality is uncertain.

Consider a hypothetical B2B software company. Its prior campaign data suggests that each additional 10,000 qualified search impressions has historically been associated with:

320 site visits
16 demo requests
4 qualified opportunities
1 closed-won customer
€24,000 in first-year recurring revenue

If a proposed budget shift is expected to add 30,000 qualified impressions, the planning case might estimate 960 visits, 48 demos, 12 opportunities, and 3 wins, or €72,000 in first-year recurring revenue. This is a scenario built from that company's history. It is not a market benchmark.

A disciplined forecast would add a low case and a high case, account for sales capacity, and state the lag between visibility and pipeline. It would also compare incremental gross profit with incremental spend. SOV is the allocation signal. Unit economics decide whether the shift deserves approval.

For a revenue operations benchmark, start with your trailing conversion rates for the same channel, offer, market, and buyer segment. If those rates are unstable, report a range rather than a point estimate. A borrowed industry average can be useful as a reasonableness check, but it should not run the forecast.

A practical brand AI visibility floor example

Averages hide weak buying situations. An AI visibility floor finds the weakest strategic query cluster.

Take a fictional B2B company called NorthstarOps. It sells planning software to operations leaders. The team tracks 50 prompts across five clusters:

Category discovery
Vendor comparison
Integration requirements
Security and procurement
Switching from a legacy system

The prompts run across three named AI systems on a fixed monthly schedule. Each cluster therefore has the same planned number of observations. After exclusions for failed responses, the team calculates recommendation rate by cluster:

Category discovery: 26%
Vendor comparison: 18%
Integration requirements: 11%
Security and procurement: 8%
Legacy-system switching: 3%

The average may look tolerable, but the AI visibility floor is 3%. NorthstarOps is almost absent from answers used by buyers considering a switch. That cluster also maps to late-stage demand, so the gap deserves attention before the team spends more on broad category awareness.

The team then inspects the actual answers. Competitors are recommended because third-party pages clearly compare migration effort, data import, implementation time, and customer support. NorthstarOps has product documentation, but little independent evidence and no concise migration page that answers those questions.

The response plan is specific:

Write a factual migration resource with supported claims and clear limitations.
Seek relevant third-party reviews and practitioner coverage that discusses switching criteria.
Improve documentation structure so answer systems can retrieve exact product facts.
Repeat the fixed prompt set and track recommendation rate, citation share, and referral activity.

This work sits within Answer Engine Optimization, where the operating goal is to make a brand understandable, credible, and retrievable for the questions buyers ask AI systems.

An AE or SDR can add commercial context during discovery. Useful questions include:

Which tools did you ask an AI assistant or search engine to compare before this call?
What buying question were you trying to answer?
Which vendors appeared repeatedly?
Did the answer cite a review, vendor page, analyst source, or community discussion?
What concern moved you from research into a sales conversation?
Which requirement removed a vendor from consideration?
Who else will check or repeat the research before a decision?

These questions should feed a structured field, not disappear into call notes. They help marketing compare measured AI visibility with real buying behaviour. They also expose pipeline leakage when a brand enters discovery late or gets excluded before an SDR ever sees the account.

What “good” share of voice looks like

There is no universal good SOV percentage. A 12% share can be strong in a fragmented category with 40 credible brands and weak in a category where four vendors control most demand.

Set a benchmark through five comparisons:

Your own stable baseline across several matching periods.
Your share of market, using the same category boundary where possible.
The median and leading competitor within the frozen peer set.
SOV by buying situation, keyword cluster, audience, geography, or product line.
The conversion efficiency of the visibility you gained.

Excess share of voice can be written as:

ESOV = share of voice - share of market

If SOV is 18% and share of market is 12%, ESOV is positive 6 percentage points. That may support a growth hypothesis. The right next step is to test whether the added visibility reaches category buyers and converts efficiently.

Benchmark quality also depends on competitor-set design. A practical peer set often includes direct sales competitors, a category leader, and one emerging alternative that appears in buyer research. The exact count matters less than stability and relevance. Five to eight brands is often manageable for an operating dashboard, but this is a working heuristic rather than a research finding. A fragmented consumer category may require a larger set.

Log competitors outside the frozen set in a watchlist. Review that watchlist at the reporting-period boundary. If a new rival must be added immediately, restate prior periods using the new set or begin a clearly labelled series. Quietly changing the denominator corrupts the trend.

Who should use SOV, and where it is less effective

Share of voice works best for teams that have a defined category, observable competitors, enough activity to reduce noise, and a recurring budget decision tied to visibility.

Strong use cases include:

SEO teams comparing visibility across stable commercial keyword groups.
Paid media managers diagnosing budget-limited and rank-limited impression loss.
PR teams comparing relevant coverage rather than counting all mentions.
Brand teams relating visibility to share of market and branded demand.
Competitive-intelligence teams tracking where rivals appear in buyer research.
Revenue teams testing whether visibility gaps align with missing pipeline segments.
AI marketing teams measuring recommendations and citations for controlled buyer questions.

SOV is less effective as the main KPI for:

A narrow experiment with a tiny number of impressions or mentions.
A new category with no stable competitor set.
A single-account campaign where buying-group engagement matters more than category visibility.
A product with long, irregular buying cycles and sparse conversion data.
A crisis period where mention volume rises for the wrong reason.
A market where brand names are highly ambiguous and cannot be cleaned reliably.

Use this diagnostic before adopting SOV:

Can you obtain comparable data for your brand and relevant competitors?
Do you need a visibility-share measure rather than a conversion KPI?
Is the sample large and stable enough to support a trend?

If any answer is no, use a more direct metric or treat SOV as exploratory. SOV fails in weaker contexts because denominator noise becomes larger than the business signal.

Three measurement mistakes that ruin the report

1. Changing the competitor set mid-period

Problem: A team adds or removes competitors during the quarter, then compares the new percentage with the old one.

Why it matters: SOV is a fraction. Changing the denominator can move the result even when the brand's own visibility stays flat.

Quick fix: Freeze the peer set for the period. Keep a dated change log and apply approved changes at the next cycle. If an urgent change is required, restate the baseline.

2. Comparing incompatible units

Problem: The executive dashboard compares search impression share, social mention share, article count, and AI recommendation rate as if one point means the same thing everywhere.

Why it matters: Each channel has a separate numerator and business question. A 20% social mention share and 20% AI recommendation share are numerically similar but operationally unrelated.

Quick fix: Label every metric with its numerator, denominator, scope, and time window. If an executive index is required, normalize the series and disclose the weights.

3. Treating all observations as equally valid

Problem: Bots inflate social mentions, syndicated press articles multiply one story, ambiguous names create false positives, and repeated AI answers are treated as independent evidence.

Why it matters: A larger count can reflect measurement artefacts rather than stronger visibility.

Quick fix: Write exclusion rules before data collection, sample-check classifications, track duplicate rates, and keep raw counts beside cleaned counts. For AI measurement, use repeat runs and retain answer-level records.

A useful check: if an analyst cannot recreate last quarter's SOV from saved inputs and written rules, controlled measurement is probably not happening.

A four-step implementation plan

Step 1: Write the measurement contract

Define the business question, channel, numerator, denominator, peer set, source, date range, market, language, device, query set, exclusions, and owner. Give the contract a version number. Any changed field starts a new version or requires a restated history.

Step 2: Build a channel data layer

Collect channel-native records before calculating percentages. Search records need keyword, rank or impression data, device, and location. Social records need mention text, source, author type, and exclusion status. PR records need publication, article URL, date, relevance, and duplication status. AI records need prompt, model, run time, answer, recommendation status, citation, and query cluster.

Step 3: Create the SOV dashboard

The dashboard should include current SOV, absolute percentage-point change, relative change, numerator volume, denominator volume, data-quality flags, and downstream conversion measures. Use a trend line for each channel, a competitor comparison, and a cluster view for buying situations. Add visible annotations when campaigns, market events, or methodology changes affect interpretation.

Step 4: Run a 90-day activation cycle

During days 1 to 30, establish the baseline and audit data quality. During days 31 to 60, run one or two controlled interventions in the weakest commercially relevant area, such as a search content update, a paid bid adjustment, a PR evidence campaign, or an AI citation project. During days 61 to 90, compare movement against the frozen baseline and inspect pipeline effects. Continue, revise, or stop based on incremental economics.

Copyable spreadsheet and dashboard specification

A useful SOV spreadsheet needs enough detail for another analyst to reproduce the result. Create one workbook with separate sheets for configuration, raw channel records, calculated metrics, change log, and revenue model.

The configuration sheet should contain:

Reporting period
Brand name and accepted variants
Frozen competitor names and variants
Geography and language
Channel definitions
Query, keyword, or publication universe
Exclusion rules
Data sources
Metric owner
Method version

The raw-data sheet should use one observation per row. Include these common columns:

Observation date
Channel
Source
Brand
Competitor-set version
Query or topic cluster
Raw visibility value
Qualified visibility value
Exclusion reason
Notes

The calculation sheet can use these fields:

Your qualified visibility
Peer-set qualified visibility
Total qualified visibility
Current SOV
Prior-period SOV
Percentage-point change
Relative change
Sample size
Data-quality warning

Use these spreadsheet formulas conceptually:

Current SOV = your qualified visibility / total qualified visibility

Percentage-point change = current SOV - prior SOV

Relative change = (current SOV - prior SOV) / prior SOV

Handle a zero prior value explicitly. Relative growth from zero is undefined, so report the new percentage and numerator instead of an infinite growth figure.

The revenue-model sheet should link the visibility change to observed funnel rates. Include qualified impressions or mentions, visits, responses, leads, opportunities, wins, revenue, gross margin, spend, and confidence range. Do not hard-code an assumed SOV-to-revenue multiplier.

A monthly operating report can follow this sequence:

State scope and method version.
Report channel SOV with absolute and relative movement.
Explain denominator changes and data-quality issues.
Identify one commercially relevant gap.
Propose a budget action with expected cost and funnel assumptions.
Name the owner and decision date.
Record what would cause the team to stop the action.

That final item protects the budget. A SOV initiative should have a stopping rule, just like a demand-generation test.

How to turn SOV into a budget decision

A budget shift should connect a visibility gap to a plausible mechanism and a measurable outcome.

Use this decision sequence:

Find a material gap within a stable dataset.
Confirm that the gap exists in a buying situation tied to revenue.
Identify the likely cause, such as insufficient bids, weak rankings, missing editorial evidence, or poor AI citation presence.
Estimate the cost of closing part of the gap.
Apply your historical conversion rates and margin assumptions.
Run a limited test before moving the full budget.

Suppose organic search SOV is already 34% in informational queries but only 9% in integration-related commercial queries. Paid search has 62% impression share for those same integration terms, with most lost share attributed to budget. AI answers recommend the brand in only 6% of integration questions, and they frequently cite competitor documentation.

The sensible response is not a broad brand campaign. The team could improve integration pages, increase paid coverage on selected terms, and earn credible third-party references for the integrations buyers care about. Each action receives its own owner and measurement path.

Budget should follow the constrained point in the buying path. If more visibility produces low-quality traffic, the constraint may be targeting or offer fit. If qualified opportunities rise but wins do not, sales execution, product gaps, or stakeholder coverage may be the problem. SOV can find an exposure deficit. It cannot repair every source of pipeline leakage.

Tactical FAQs about share of voice

How often should I measure share of voice?

Measure paid search and fast-moving social channels weekly for operational control, then report decisions monthly. SEO, PR, and AI recommendation SOV often work well on a monthly cadence, with quarterly method reviews. Use longer windows when volume is low or seasonality is strong.

Can I compare SOV across channels?

You can compare direction and strategic gaps, but raw percentages do not have equal meaning. Keep channel-native figures separate. A normalized executive index is acceptable if the weights and method remain visible.

How many competitors should I track?

Track the smallest set that reflects real buyer choice. Five to eight brands is a manageable starting heuristic for many B2B categories. Include direct rivals and any alternative that repeatedly appears in search results, AI answers, reviews, or sales discovery. Freeze the set for the reporting period.

What happens when a new competitor enters mid-quarter?

Add it to a watchlist first. Bring it into the main set at the next reporting boundary. If immediate inclusion is necessary, recalculate prior periods with the new denominator and label the restated series.

How does share of voice relate to conversions?

SOV measures relative visibility. Conversion data measures what happened after exposure or engagement. Relate them through your own history by channel and buyer segment. Avoid claiming that a fixed SOV gain creates a fixed pipeline lift.

Is share of voice the same as Google Ads impression share?

No. Google Ads impression share uses eligible impressions as the denominator. Competitive SOV commonly uses the measured visibility of a peer set. Paid reports should name the exact metric rather than using “SOV” as a catch-all label.

How should I handle bots and duplicate mentions?

Create exclusion rules before reporting. Remove obvious automation, spam, duplicated syndication, irrelevant name matches, and campaign-generated repetition when it does not reflect the intended question. Save raw and cleaned counts so the adjustment is auditable.

Can AI share of voice be measured reliably?

It can be measured directionally with a fixed prompt set, named models, repeat runs, answer-level records, and clear recommendation rules. It should not be presented as a census of everything an AI system says. Model updates and answer variation require repeated observations.

Should branded queries be included in SEO SOV?

Report branded and non-branded queries separately. Branded queries measure existing demand and brand access. Non-branded category queries measure visibility before the buyer has selected a brand. Mixing them can make organic performance appear stronger than it is during discovery.

What should I do if SOV rises but pipeline stays flat?

Check traffic quality, query intent, audience fit, offer conversion, attribution lag, SDR follow-up, opportunity qualification, and multi-threading. The visibility may be real but commercially weak. It may also be too early for the chosen sales cycle. Hold the budget increase until the team can explain the break in the path.

The 2026 outlook: AI visibility joins the operating report

In 2026, search rankings remain useful, but they no longer describe the full discovery process. Buyers can ask an AI system for a shortlist, a comparison, implementation risks, or alternatives before visiting a vendor site. A brand may rank well in search while appearing rarely in generated answers.

This changes the workflow in several practical ways.

First, prompt sets become managed research assets. Teams need version control, buying-stage labels, model coverage, and repeat-run rules. A random list of prompts produces screenshots, not a decision system.

Second, citations matter alongside mentions. AI systems often rely on product pages, documentation, reviews, news coverage, forums, and other third-party sources. Marketing needs to know which sources recur, where product facts are absent, and which claims lack credible support.

Third, sales discovery becomes a source of query intelligence. SDRs and AEs hear how prospects frame problems, compare vendors, and test risks. Those phrases should feed the prompt library. Marketing can then measure whether the brand appears in the same questions that produce pipeline.

Fourth, revenue attribution will remain probabilistic in many cases. Referral traffic, self-reported attribution, call notes, CRM fields, and controlled query monitoring can be combined, but none provides a perfect census. Teams should state confidence levels rather than force false precision.

The management standard will rise. CMOs will need separate reporting for search visibility, paid eligibility, social conversation, editorial presence, AI recommendations, and citations. The teams that maintain stable definitions will learn faster than teams that chase every weekly score movement.

The point of the system

Share of voice will not prove causation, fix weak positioning, or guarantee revenue. It will tell you where your brand is present, where competitors have more visibility, and whether a controlled intervention changed that position.

That is enough to improve budget discipline. Keep the peer set fixed. Keep channel units separate. Report percentage-point and relative change. Tie the investment case to your own funnel history. Then move money toward a measured constraint rather than the loudest internal request.

Targetlytics helps revenue-driven teams measure AI recommendations, citations, competitors, and the buyer questions behind generated answers. You can start with a free AI visibility audit and book a call. You can also start for free, with paid plans offering a 14-day trial when you are ready to run the workflow continuously.