Skip to content
Client login

Free Audit

Digital marketing

The arithmetic of a rating: why volume is the only lever

A one-star review is a catastrophe at twelve reviews and a footnote at two hundred. What that implies, and what to do instead of panicking.

Published on 2 August 2026 — Algeria Agency

The article accompanying this one deals with the listing: claiming it, what anybody can edit, the photographs, the duplicates, what can be reported, how to reply. It deals with the object.

This page deals with the number displayed on it, and with a property of that number almost nobody holds in mind: a mean behaves differently depending on how many values it aggregates, and that difference decides everything else.

The practical consequence is blunt. At twelve reviews, one one-star review costs several tenths of a point and shows in the search result. At two hundred, the same review moves the mean by a hundredth and nobody sees it. This is not a question of severity, it is a division.

So this article is about what that implies: why volume is the only real lever, what it actually takes to raise a rating, when to ask, what must never be done, and why a perfect score is a problem rather than a target.

A mean is not an opinion, it is a division

The starting point is arithmetic nobody sets out, and it explains almost all business behaviour around reviews.

Take a business with twelve reviews, all five stars: the rating is 5.0. A one-star review arrives: it becomes 4.7. The same review, at a business with two hundred five-star reviews, moves the rating from 5.0 to 4.98 — displayed as 5.0. The content is identical, the effect is incomparable.

The same arithmetic works the other way, and that is where it becomes useful. A business at 3.8 over twenty reviews rises to 4.2 with ten new five-star reviews. The same business at 3.8 over two hundred needs roughly a hundred five-star reviews for the same shift.

What to retain is not a figure but a shape: every new review weighs less than the one before. A small number of reviews makes a rating volatile, a large number makes it stable, and stability protects in both directions.

From this follows the only strategy that genuinely exists in this area, and it is almost a disappointment for how simple it is: collect reviews regularly, in volume, indefinitely. Everything else — replying, reporting, worrying — does not act on the count.

What the rating actually does to a decision

Before putting effort into it, you need to know what this number changes, and the answer is more nuanced than "you should be well rated".

The rating works first as a coarse filter rather than a fine ranking. People rule out what is below a threshold and do not discriminate finely above it. The difference between 4.3 and 4.6 weighs little; the difference between 3.4 and 4.3 is decisive.

The number of reviews acts as a second filter, and it acts independently. An excellent rating over four reviews is read as an absence of information, sometimes as suspicion. Conversely, an average rating over three hundred reviews reads as reliable information, and it is more reassuring than a perfect score over six.

What people read next is not the mean but the most recent reviews, and that changes the priority. Three negative reviews from two years ago weigh less than two from this month, whatever the mean — because the reader wants to know how it is now.

The practical conclusion is two points: getting out of the low threshold is urgent, and gaining two tenths above four is not. The regularity of recent reviews matters more than the displayed value.

The only lever: asking, regularly

Almost every satisfactory business has few reviews, and the reason is not dissatisfaction: it is that happy people do not think to write. Dissatisfaction motivates spontaneously.

That imbalance produces most unfairly low ratings. A business serving two hundred people a month well and asking none of them mostly receives reviews from those who had a problem — three a quarter, all negative, on a tiny base.

Asking corrects that bias, and it is the entire mechanism. It is not about obtaining good reviews but representative ones, which mechanically produces a higher rating if the service is sound — and a justly low one if it is not, which is information you want to have.

Regularity matters more than intensity. Five requests a week for a year produce a solid base and a flow of recent reviews; fifty requests in one week produce a visible anomaly and nothing afterwards.

So set a modest, sustainable number attached to an existing operation rather than to good intentions: at every delivery, at the end of every job, at every till transaction on a given day.

The moment, and the person who asks

The request works or does not according to two parameters, and neither of them is the wording of the message.

The first is the moment. It sits just after the customer has got what they wanted, while the satisfaction is present — at handover, at the end of the job, when the problem has just been solved. A request sent three days later lands in a day that has nothing to do with you.

The second is the person. A request made by whoever did the work gets far more responses than an automated message, for a simple reason: the customer answers somebody, not a business. "If you were happy with it, it would help me a lot" beats any professional formulation.

Friction decides the rest. Every additional step divides the number of responses: searching for the business, finding the button, logging in. A short link or a visual code that opens the form directly is what separates a request that works from a polite one.

One local point not to neglect: many people here are used to messaging and little to forms. Sending the link through the messaging channel where the conversation is already happening produces noticeably more responses than the same link by another means.

What must never be done

Three practices circulate, they look effective, and each costs more than it returns — the third potentially costing the whole listing.

The first is offering something in exchange: a discount, a gift, entry to a draw. It is against the platform’s rules, it is detectable in the text of reviews, and it distorts what you are measuring. A bought review teaches you nothing about your service.

The second is asking only the customers you expect a good rating from. It is very widespread and also against the rules. But the decisive argument is elsewhere: it removes precisely the information you need, and a business that filters its requests stops knowing what is wrong with it.

The third is buying reviews. The accounts used are identifiable, waves are detected, and the penalty falls on the listing — mass removal of reviews including legitimate ones, or suspension. It is the only practice on this list that can destroy in one go what three years built.

There is a fourth, less serious and very common: asking employees and family. Those reviews are spotted — same week, accounts with no history, similar phrasing — and they have the drawback of making the genuine reviews around them look suspect.

Replying addresses the next reader

A reply to a review is not there to convince its author, who has rarely changed their mind, but to be read by the people who come afterwards — and they are far more numerous.

That changes what to write. A reply that rebuts point by point is addressed to the author; a reply that briefly explains what happened and what was done is addressed to the reader, and it is the one that works for you for two years.

Length is a signal in itself. A long reply to a short review gives the impression of a defensive business, whatever its content. Three lines is almost always enough, and the length constraint mechanically improves the tone.

Reply to positive reviews too, briefly and without a repeated formula. This is not politeness: a listing where each review has a specific reply shows a business that reads, which is exactly what somebody hesitating is looking for. Twelve instances of "Thank you for your trust!" produce the opposite effect.

One absolute rule: never reply the same day to a review that angers you. The delay costs nothing — a review is not an incident, it does not spread — and the reply written the next day is consistently better.

When the rating is low and deserved

This case has to be planned for because it is frequent and because everything above becomes counter-productive when it applies.

The sign is recognisable: the negative reviews all say the same thing. Three people mentioning the same delay, the same salesperson, the same fault are not describing three bad experiences but an operational problem that produces reviews.

In that case, asking for more reviews makes it worse. Collection is an amplifier: it produces a representative rating, and if the service is poor the representative rating is poor. A business that increases its review volume before fixing its problem simply accelerates the arrival of the truth.

So the order is the reverse of what everybody applies: fix first, collect afterwards. And the useful point is that the reviews tell you exactly what to fix, for free, with dates — which makes them the best customer feedback a small business has.

Once the problem is corrected, collection resumes its role and works quickly, because new reviews describe the new situation and readers look at the most recent. It is the only case where a rating visibly recovers within months.

Old reviews that anchor

A rating contains reviews from every period, and a business that has changed drags what it used to be for a long time. It is the commonest frustration in this area and it has an answer, if a slow one.

What does not work: asking for an old review to be removed because it no longer describes reality. That is not an admissible ground, and age is not one either.

What partly works: replying to the old review today, dating the reply. "This comment is from 2024; we have since changed supplier and the lead time is now three days" is read by everybody who reads the review, and it turns criticism into a demonstration of follow-through.

What genuinely works is the flow: readers look at recent reviews, and an accumulation of consistent recent ones pushes the old ones down visually and statistically. This is section 1 again — volume is the lever, here too.

One exception is worth knowing: an old review can be removed if it breaches the rules — abuse, unrelated content, an obvious conflict of interest. Those are the grounds in the corresponding section of the neighbouring article, and they do not depend on the date.

A perfect score is a problem

Many businesses aim for five out of five, and it is a counter-productive target worth understanding.

A perfect score over a significant volume reads as improbable, and it is: any activity serving hundreds of people produces dissatisfied ones. The reader infers either a small volume or filtering, and both reduce the trust they place in the whole.

Middling reviews in fact play a useful role nobody credits them with. A well-written three-star review, with a measured reply from the business, is often the most persuasive element on a listing — it proves the reviews are not filtered and it shows how you react when things go wrong.

The comfortable zone therefore sits lower than people think, and it is paired with a volume: a rating between 4.3 and 4.7 over several dozen reviews, with recent ones, is more solid than a 5.0 over twelve.

That frees you from two expensive behaviours: panic after an isolated negative review, and effort spent trying to remove three-star reviews, which are generally the most useful items on the listing.

What to measure, once a quarter

Looking at your rating every day serves nothing — it is a mean, it barely moves, and looking at it mostly produces anxiety. Four figures recorded once a quarter are enough.

The first is the total number of reviews, because it is the only quantity you act on directly. The second is the number received this quarter, which says whether collection is still working or has stopped without anybody noticing.

The third is the rating of this quarter’s reviews alone, separated from the overall rating. It is the most useful figure on the list and it is invisible on the listing: it describes your service now, whereas the mean describes your history.

The fourth is the subject of the negative reviews, grouped. Three reviews mentioning the delay are not three reputation problems, they are an operational problem — and that grouping is the only reliable link between this activity and the rest of the business.

Record these four figures in the same document as the previous quarter. A series of four quarters shows a trend no daily checking reveals, and it fits on four lines.

When there are no reviews at all

This is where many businesses start, and it is less bad than a low rating, provided you do not stay there — because a listing with no reviews reads as a business nothing is known about.

The first target is modest and precise: about ten reviews. Below that the displayed rating means nothing and readers sense it. Above it, the listing starts to look like that of a business that exists.

The quickest route is asking recent customers rather than old ones, for two reasons: they remember, and their reviews describe the current situation. A list of fifteen customers served within the month generally produces five to eight reviews if the request is made properly.

Do not ask everybody in one week. A listing receiving twelve reviews in three days and then nothing for six months is visible and suspect; the same twelve spread over two months pass unnoticed and leave a flow.

And accept that the first reviews will be uneven. A business starting at 4.1 over eleven reviews is in a far better position than one with none, even if the number disappoints — because the reader now sees information rather than a void.

What we do, and what we refuse

What we do is bounded. We set up the review request at the right moment in your operation, we build the short link or code that removes the friction, we draft the replies, and we record the four quarterly figures with you.

We refuse to offer anything in exchange for a review, including a modest discount and including when a competitor is visibly doing it. It is against the platform’s rules, it is detectable, and the risk falls on the whole listing — that is, on years of legitimate reviews.

We also refuse to ask only the customers expected to give a good rating. It is the most widespread practice on this list and the most tempting; it removes precisely the information collection was meant to produce, and a business that filters its requests ends up not knowing what is wrong with it.

And we do not promise a rating. Regular collection produces a representative rating, which is good news when the service is good and bad news when it is not — in the second case what we recommend is fixing the problem first, because collection only accelerates the arrival of the truth. There is also something only you can do today: ask the next satisfied customer for a review yourself, at the moment they thank you.

Frequently asked questions

A one-star review has just arrived. Is it serious?

It depends entirely on your volume. At twelve reviews a one-star costs several tenths and shows; at two hundred it moves the mean by a hundredth. It is a division, not a question of severity. So the useful response is not panic but increasing the number of reviews, because that is the only lever there is.

What does it take to raise a rating?

Volume, and the larger the base the more of it. A business at 3.8 over twenty reviews rises to 4.2 with ten new five-star ones; the same at 3.8 over two hundred needs about a hundred. Every new review weighs less than the last — which makes a small base volatile and a large one stable, in both directions.

When should we ask for a review?

Just after the customer has got what they wanted, and from whoever did the work — a customer answers a person, not a business. Reduce friction: a short link or code that opens the form directly, sent through the channel where the conversation is already happening. Every extra step divides the number of responses.

Can we offer a discount for a review?

No. It is against the rules, detectable in the text of reviews, and it distorts what you are measuring. Same for asking only customers you expect a good rating from: it is the most tempting practice and it removes exactly the information you need. And buying reviews can cost the whole listing, legitimate reviews included.

Should we aim for five out of five?

No, it is counter-productive. A perfect score over a significant volume reads as improbable, and the reader infers a small volume or filtering. A rating between 4.3 and 4.7 over several dozen recent reviews is more solid. Well-answered three-star reviews are often the most persuasive items on a listing.

Our negative reviews all say the same thing. What now?

Fix the problem before collecting. Collection is an amplifier: it produces a representative rating, so if the service has a real fault, raising the volume only accelerates the arrival of the truth. Three reviews mentioning the same delay are not three reputation problems, they are an operational one — reported to you for free, with dates.

Where we come in

The arithmetic is simple and reassuring: volume dilutes everything. It does not say at which moment of your operation the request actually gets answered.

  • We put the request where the customer thanks you, not three days later.
  • We make the physical carrier: a card, a notice, a short link you can say aloud.
  • We take the rate every month and tell you when it is slipping.

We will never ask only your happy customers: the collection covers everyone, or it is worth nothing at all.

Read next

Let us talk about your project

A free audit, no commitment: we look at your online presence and tell you what is holding it back.

We measure how this site is used with Google Analytics, to learn which pages actually help. You can stop that measurement at any time from the footer. Cookie policy