How to Measure Sampled Social Share of Voice

Measure sampled social share of voice by fixing the comparison rules, collecting inspectable evidence and applying an explicit counting convention. The arithmetic is simple; the difficult part is preventing unrelated names, duplicate posts and changing source coverage from creating a misleading result.

Write the research protocol before collecting

Name the brands, aliases, platforms, language, intended period and topic. Define whether own-brand posts and sponsored content are included. Specify what counts as an eligible mention, how multi-brand posts are treated and whether the report counts posts or brand-post incidences. Use the share-of-voice definition guide to choose the denominator.

Avoid asymmetric queries that use a broad category for one brand and an exact product name for another. Record ambiguous aliases and product context. A brand named after a common noun needs relevance checks, not an assumption that every match is a brand mention.

Build and clean the source ledger

• Save original source URL, platform, post date and observation date.
• Assign eligibility: relevant brand/context, unrelated name, duplicate or unavailable for review.
• Record brand mentions using the same rule for every competitor.
• Mark sponsored/owned content if those distinctions affect your question.
• Preserve exclusions and the reason for each, rather than deleting inconvenient examples without a trace.

Brand24’s Mentions documentation describes filtering, source links and removal of irrelevant items. Meltwater’s monitoring description describes competitor and historical analysis. These references establish practical capabilities to evaluate; they do not prove complete collection for your query or define every vendor’s share formula.

Worksheet: make the calculation reproducible

Worked example: one duplicate changes the split

An invented ledger initially contains 20 eligible brand-post incidences: A has 8, B has 7 and C has 5. Under the declared incidence convention, their shares are 40%, 35% and 25%. Review finds that two A records point to the same original post and one should be removed. The corrected counts are 7, 7 and 5, total 19; A and B each have about 36.8%, while C has about 26.3%.

The correction changes the interpretation from “A leads” to an equal observed count for A and B. The team preserves the excluded URL and notes the correction. It then reads the original posts: some describe support problems and others product demonstrations. The split does not establish which brand is commercially stronger.

Use Superlurk’s controls accurately

The Share of Voice tool accepts a subject and named competitors. Keep your date and eligibility requirements in the research protocol, then inspect whether the evidence satisfies them. Review the output’s own counting explanation rather than assuming it implements the illustrative convention above. Tracking in the current workspace uses manual checks.

Compare periods with appropriate limits

Keep rules sufficiently stable, but also log changes in access, query wording, ranking and available sources. Even an identical query does not guarantee an identical collection process. Report the count, denominator, relevant examples and gaps alongside any percentage. Investigate large changes before attributing them to a campaign.

Use competitor analysis to turn the sources into a proposed test and brand sentiment review to understand the context. A share-of-voice result is one observation in that decision, rather than a sales or audience estimate.