How do you measure reputation?
Reputation is not one number, and any product selling one is publishing its own arithmetic. What can be measured honestly is five separate things: the star rating on the profiles people actually read, the velocity at which new reviews arrive, the composition of the first page of results for your name, the sentiment of what is written about you, and the volume of branded search. Each answers a different question and each has a specific blind spot. This page sets out what every measure shows, what it misses, and how to fix a baseline you can compare against next quarter.
Star rating: the number most people see first
The star average on a Business Profile is the most visible reputation measure that exists, and it is also the crudest. It compresses every experience into one figure, it is slow to move once volume is high, and it says nothing about why it is what it is.
Read it in three parts rather than as one number. The average tells you the headline. The count tells you how much weight the average deserves, because an average built from twelve reviews is a different object from one built from twelve hundred. The distribution tells you the shape: a profile sitting at four stars because most reviews are four is a different business from one at four stars because the reviews are split between five and one.
Star rating and review count are also inputs to local ranking, which Google states directly in Google's guidance on local ranking factors, so the number is doing work beyond persuasion.
Review velocity: how recent the evidence is
Velocity is the measure most people skip and the one that changes fastest under effort. Track new reviews per month across every platform you hold, and track it as a line rather than a total, because the shape carries the information. A steady line is healthy. A flat line with occasional spikes usually means somebody runs a review campaign when the rating dips, which is the arrival pattern automated filters treat as suspicious.
Velocity is also the early warning. A drop in new reviews shows up months before it shows up in the average.
SERP composition: what the first page is made of
For anything larger than a single storefront, the first page of search results for the brand name is the reputation. Measuring it means counting rather than reading.
Search the name in a signed out window, take the top ten results, and classify each one: owned property, controlled profile, neutral third party, or negative. Record the classification and the position, and repeat it on a fixed schedule. What you get is a composition figure you can compare against last quarter, rather than an impression.
The same exercise on the image results and on the autocomplete suggestions is worth doing once a quarter. Both are visible to every searcher and neither appears in any dashboard.
Sentiment: what the writing says, at scale
Sentiment analysis assigns a polarity, positive, negative or neutral, to a piece of text, so that a body of reviews or mentions too large to read can be summarised. It is genuinely useful at volume and genuinely unreliable at the level of the individual item.
The failure modes are well documented in the research literature. Sarcasm, negation, comparison, industry jargon and very short texts are all hard, and a model trained on general text does poorly on domain language. The ACL Anthology of natural language processing research is where the field publishes its own results, including the error rates, and it is a useful antidote to a vendor dashboard that reports sentiment to one decimal place.
Use sentiment as a trend and a triage tool. Treat any single classification as a suggestion, and read the underlying text before acting on a change.
Branded search volume: whether the name is being looked up
Branded search volume is the number of people searching your business name. It is a demand measure rather than an opinion measure, and it belongs here for two reasons. It tells you how many people are exposed to whatever the first page contains, which is what turns SERP composition from a curiosity into an exposure figure. And a sustained fall in branded search after an incident is one of the few reputation effects that shows up as a number rather than as a feeling.
You can read it from your own search analytics for queries containing the name, and from any keyword tool for the name itself. Both are approximations. Watch the direction, not the absolute value.
The reputation metrics side by side
| Measure | The question it answers | Where it comes from | What it misses |
|---|---|---|---|
| Star rating | What does a stranger see first | Business Profile and other review platforms | Why the number is what it is; slow to move at high volume |
| Review velocity | Is the evidence current | Count of new reviews per month, per platform | Nothing about content or sentiment |
| SERP composition | What does page one for the name contain | Manual signed out search, classified and counted | Personalisation and location shift results between observers |
| Sentiment | What is the tone of what is written | Text analysis over reviews and mentions | Sarcasm, negation, jargon, short texts; unreliable per item |
| Branded search volume | How many people are looking | Search analytics and keyword tools | Intent behind the search; approximate by nature |
Setting a baseline you can compare against
A measure taken once is an anecdote. The value is in the comparison, which means fixing the method before the first reading.
- Write down the method, including the exact search terms, the platforms counted, and whether searches are run signed out.
- Take the first reading and date it. Screenshots for anything visual, numbers in a sheet for the rest.
- Fix the interval. Monthly for velocity and star rating, quarterly for SERP composition and branded search.
- Change one thing at a time between readings, so a movement can be attributed.
- Keep the raw record, not only the summary. A dashboard that recalculates history is not evidence of what was there.
Which reputation metrics matter most
It depends on what the business actually depends on, and being honest about that beats tracking everything. A local service business lives on star rating and velocity, because a Business Profile is the first and often the only thing a customer reads. A professional or an executive lives on SERP composition, because the decision that matters is made after someone searches a name. A larger organisation with real mention volume gets value from sentiment and share of voice that a small business cannot, because the sample is too small to be stable at low volume.
One measure not on this list is any single number sold as an overall reputation score. There is no universal standard behind one, and the products that publish a score are calculating their own from inputs they chose.
The other thing worth saying plainly: the way not to move any of these numbers is to buy the inputs. Incentivised and fabricated reviews are the subject of a federal trade rule, and the FTC's endorsement guides set out the disclosure expectations that go with endorsements generally. A metric moved that way measures the purchase, not the reputation.
Establishing the first baseline across all five is most of what a reputation audit produces.
Questions about how to measure reputation
How do you measure reputation?
With several measures rather than one: star rating and review count, review velocity, the composition of page one for your name, sentiment across reviews and mentions, and branded search volume. Fix the method first, then compare readings over time.
What metrics matter most?
It depends what the business depends on. Local service businesses live on star rating and velocity. Professionals and executives live on what page one for their name contains. Sentiment and share of voice only become stable at real mention volume.
Is there a single online reputation score?
Not a universal one. Products that publish a score are calculating it from inputs they chose, using weightings they set, so the number is only meaningful inside that product and cannot be compared across tools.
How often should reputation be measured?
Monthly for star rating and review velocity, quarterly for SERP composition and branded search volume. More frequent readings mostly capture noise, and less frequent ones miss the point at which something changed.
Can sentiment analysis be trusted?
As a trend across a large body of text, reasonably. On a single review, no. Sarcasm, negation, comparison and industry jargon are all documented weak points, so a change in the sentiment line is a prompt to read the underlying text, not a finding on its own.