Most advice about online reviews reduces to “get more of them and make them good.” True, unhelpful, and it obscures the most commercially useful finding in the entire dataset.
Review impact is not linear. It has thresholds, and the return on effort is wildly different depending on which side of one you are sitting.
The finding
BrightLocal’s analysis of conversion by star rating produced two results that sit oddly together.
Conversion rates peak at around 4.9 stars. Unsurprising.
But the largest growth in conversion came from businesses that moved from 3.5 stars to 3.7 stars over a year, a jump of nearly 120%, the highest percentage growth from any rating change.
Two tenths of a star producing more than a doubling in conversion.
This does not fit a model where each additional half-star adds proportional value. It fits a model with a threshold, and the threshold appears to sit somewhere around 3.6.
Why a threshold exists
The mechanism is about how people use a rating, which is not as a measure of quality but as a filter for risk.
A rating below roughly 3.5 reads as a warning. Something is wrong, and the reviews will tell you what. Most people do not investigate. They pick a different option, because the cost of doing so is zero.
A rating in the high 3s reads as normal. Not exceptional, but not a red flag. The customer stops filtering and starts evaluating the things they actually care about: price, distance, opening hours, whether the photographs look right.
Crossing that boundary does not make you better. It moves you from the excluded pile to the considered pile, and the excluded pile converts at approximately zero.
This is why the marginal return on review work is highest for businesses that are currently mediocre. A business at 3.5 has more to gain from two tenths of a star than a business at 4.6 has from four tenths.
Attention follows rating too
Within the local pack, ratings govern how much attention a listing receives before anyone reads anything. Listings displaying five stars capture around 69% of attention, four stars around 59%, and three stars around 44%.
So a poor rating costs twice: it reduces the proportion of people who consider you, and it reduces the proportion who look at your listing at all.
Meanwhile 87% of consumers read online reviews for local businesses, which means this filter is running for nearly every customer. It is not a segment of cautious shoppers. It is the default behaviour.
Businesses with fewer than ten reviews or an average below 4.0 face a measurable conversion penalty on both counts.
Volume, recency and velocity are three different things
Businesses conflate these and optimise the least important one.
Volume establishes credibility. Ten reviews is materially different from two. Two hundred is not materially different from a hundred, the returns flatten quickly, and past a certain point additional volume mostly stabilises the average against a bad month.
Recency is what customers actually read. Almost nobody scrolls to reviews from three years ago. A business with 180 reviews averaging 4.6, whose most recent review is fourteen months old, reads as declining regardless of what the number says. The rating describes a business that may no longer exist in that form.
Velocity is the flow. A steady trickle of recent reviews signals an operating business with current customers. It also protects the average: when one bad review arrives, a business receiving four reviews a month absorbs it in weeks, while a business receiving four a year carries it for a long time.
The practical consequence is that a business should optimise for a small, consistent flow rather than for periodic campaigns. Six reviews a month indefinitely beats seventy in one quarter and nothing after.
Getting reviews without being irritating
The gap between businesses that accumulate reviews steadily and those that do not is almost never about customer satisfaction. It is about whether asking is built into the process.
Ask at the moment of relief. Not at the transaction, at the point where the customer’s problem has visibly been solved. The plumber’s moment is when the leak stops. The consultancy’s is when the visa is approved. That is when someone will actually write something warm, and it is frequently days after the invoice.
Make it one tap. Google Business Profile generates a short review link. Put it in the WhatsApp message, the invoice footer, the email signature. Every additional step loses a large share of people, and “search for us on Google and leave a review” is several steps.
Ask in person, follow up in writing. A verbal request from the person who did the work, followed by the link, converts several times better than a cold automated message. The automation is the reminder, not the ask.
Do not filter. Sending happy customers to Google and unhappy ones to a private feedback form violates Google’s policies and, more practically, produces a rating profile that reads as suspicious. A perfect five with no critical reviews is less credible than a 4.6 with a few threes.
Never buy them. Beyond the platform policy violation, fake reviews are enforceable under consumer protection law in a growing number of jurisdictions, and the detection has improved considerably. The downside is asymmetric and the upside is a number that was never the point.
Responding is half the value
Responses are read by people who are not the reviewer, which is the entire reason to write them.
Respond to everything, including the positive ones, briefly. It signals an actively operated business.
Respond to negative reviews for the audience, not the reviewer. The reviewer is generally not persuadable. The next twenty people reading are. A calm, specific, non-defensive response that acknowledges the issue and states what was done converts a liability into evidence of how you handle problems.
Never argue. A defensive response does more damage than the original review, every time. The reader is not adjudicating the dispute. They are assessing what dealing with you would be like, and the response answers that question directly.
Respond quickly. A response two days after the review reads as attentive. Six months later it reads as damage control.
Reviews as a ranking factor
Beyond conversion, reviews feed the local algorithm. Review count, average rating, velocity, and the text content all contribute to local pack ranking, and the text matters more than most businesses realise: reviews mentioning a service by name help establish relevance for that service.
You cannot script this and should not try. But asking a customer “would you mind mentioning what we did for you?” is legitimate, and it produces reviews containing the vocabulary customers actually use, which is frequently different from the vocabulary on your website.
That gap between how a business describes itself and how customers describe it is, incidentally, the most common reason a site fails to rank at all. Reviews are one of the few places you get to read your customers’ actual language, and it is worth mining them for it. Any competent seo consultant nepal or elsewhere will pull vocabulary out of review text when planning page titles, because it is free primary research into search behaviour.
Reading your reviews as research
The overlooked use of a review corpus is not reputation management. It is that reviews are the only place you get to read your customers describing you in their own words, unprompted and at scale.
Three things worth mining.
The vocabulary. Customers describe what you do differently from how you describe it. A guesthouse calls itself “boutique accommodation”; reviews say “clean rooms near the lake with good wifi.” That gap between internal language and customer language is the single most common reason a business ranks for nothing but its own name, and reviews hand you the customer half of it for free. Put those words into page titles and headings.
The recurring complaint. Almost every business below 4.0 has one specific failure appearing repeatedly, a wait time, a communication gap, one member of staff, one process. It is usually operational rather than mysterious, and fixing it is what makes the review-generation effort work. Asking for reviews while the underlying problem persists just accelerates the arrival of more bad ones.
The unexpected positive. Reviewers frequently praise something the business considers incidental. That is a positioning signal, and it is often a better hook for a homepage than whatever is currently there.
None of this costs anything and none of it requires a tool. It requires reading two years of reviews in one sitting with a notebook, which almost nobody does.
A warning on review schema
One technical note, because it is a genuine liability rather than a missed opportunity.
Do not emit AggregateRating or Review structured data with invented figures. Ratings and review counts must come from real stored submissions on your own property, and self-serving review markup on your own page breaches Google’s structured data policy, risking a manual action on a site whose entire value is organic traffic.
In several jurisdictions it is independently enforceable as a consumer protection matter, and enforcement of fake reviews and testimonials has become considerably more active.
If you want star ratings visible in search results, the legitimate routes are to collect real reviews on your own site and mark those up honestly, or to rely on the Business Profile rating which appears in the local pack without any markup on your part. A competitor visibly faking it is a record of their risk, not a template for yours.
Where this sits against everything else
For a local business, reviews are among the highest-return activities available, and the reasoning is comparative.
The local pack takes roughly 42-44% of clicks on local searches, and Google shows a local pack in about 93% of local queries. Businesses in the three-pack get 126% more traffic and 93% more conversion-oriented actions than those below it. Within the pack, ratings govern both attention and conversion.
So reviews operate at the point where the majority of local search traffic is decided, and they cost nothing but a habit.
Compare that to the alternatives at a small budget. Content production takes months and requires ongoing investment. Paid search stops when payment stops. A website redesign is expensive and does not touch the surface where local decisions are made.
This is why practitioners working on local businesses tend to sequence review work early. An seo in pokhara engagement for a single-location business, or the equivalent anywhere, will typically address the Business Profile and review flow before commissioning a single page, because the return per hour is higher and the results arrive faster.
What to do
If you are below 4.0: this is your highest-return activity, full stop. The 3.5-to-3.7 finding says the next two tenths of a star are worth more than anything else you could spend the quarter on. Read your negative reviews for the pattern (there is almost always one specific recurring failure), fix it operationally, then start asking.
If you are between 4.0 and 4.5: focus on velocity and recency rather than average. Build the ask into the process so the flow is continuous.
If you are above 4.5: protect it. Keep the flow steady so one bad month cannot move the number, and respond to everything.
In every case: make the ask one tap, ask at the moment of relief rather than the moment of payment, respond to every review within a couple of days, and never buy any of them.
Two tenths of a star. Nearly 120% conversion growth. There is no other lever in local marketing with that ratio, and it costs nothing but remembering to ask.