Not legal advice: This article covers marketing practice and general legal awareness of platform and consumer-protection rules on reviews, not a legal audit. If you run review-request campaigns, confirm current requirements — especially outside the US — with a qualified lawyer.
"Reviews are the new word of mouth" is one of those lines repeated so often nobody checks it anymore — along with a specific, confident set of numbers that ride along with it: some fixed share of shoppers who supposedly read reviews before buying, a precise revenue cost for a bad one, a precise loyalty gain for replying to one. Some of that turns out to be real, rigorously measured research. A lot of it turns out to be a number with no traceable source, repeated by marketing blogs that cite each other instead of a primary study.
We went and checked: the Harvard working paper that actually measured what a one-star Yelp change does to real restaurant revenue, the peer-reviewed study on whether a bad review costs more than a good one earns, the academic paper on what happens to your rating when you actually reply to reviews, the largest disclosed year-by-year survey of how picky shoppers have gotten about star ratings, and the rule the US FTC finalized in 2024 that makes some of the most common review-request tactics illegal outright. Several widely-quoted numbers didn't survive the check — we flag those clearly instead of repeating them.
Key takeaways
- A one-star increase in a restaurant's Yelp rating is associated with a genuine 5–9% revenue increase — but only for independent restaurants. Chain restaurants showed no statistically significant effect in the same Harvard Business School study, which used real Washington State tax revenue records, not self-reported sales.
- Negative reviews carry more weight than positive ones: a peer-reviewed 2006 study of real Amazon and Barnes & Noble sales data found a single 1-star review swung relative sales rank by an amount equivalent to about 20 fewer books sold per week, while 5-star reviews showed almost no measurable effect at one of the two retailers.
- Replying to reviews genuinely moves your rating: a peer-reviewed study of 314,776 TripAdvisor reviews across 5,356 Texas hotels found management responses were followed by a roughly 0.12-star average rating increase and about 12% more reviews, with a third of hotels seeing their displayed, rounded rating rise by half a star or more within six months.
- Google and the US FTC now both explicitly ban asking only happy customers for reviews, or paying for them: Google's policy prohibits "selectively solicit[ing] positive reviews," and the FTC's rule, effective October 2024, carries civil penalties of up to $51,744 per violation. Israel has no equivalent review-specific law on the books — a real regulatory gap, not an oversight we're inventing.
- The most commonly repeated review stat — some version of "93% of shoppers read reviews before buying" — doesn't trace to any single disclosed study; it's copied between vendor blogs that cite each other. BrightLocal's own year-by-year, disclosed-sample surveys are the most rigorous data we found, and they show star-rating expectations climbing fast: the share of shoppers requiring 4.5 stars or higher reportedly jumped from about 17% to 31% in a single year.
Does a better star rating actually increase revenue?
The most rigorous answer we could find comes from Michael Luca, a Harvard Business School economist, in a working paper titled "Reviews, Reputation, and Revenue: The Case of Yelp.com" — first posted in 2011 and revised in 2016. It's worth being precise about what this is: an HBS working paper distributed via SSRN, cited thousands of times and often described in the press simply as "a Harvard study," but never, as far as we could find, published in a peer-reviewed journal.
The data behind it is real and unusually well-matched: Yelp review and rating data covering roughly 60,000 Seattle restaurants — about 70% of the city's total — merged with actual tax revenue records from the Washington State Department of Revenue, not self-reported figures. To isolate cause from correlation, Luca used Yelp's own rounding quirk (displayed ratings round to the nearest half-star) as a natural experiment, comparing restaurants that landed just above a rounding threshold to near-identical ones that landed just below it, then ran a density test to confirm restaurants weren't gaming their way across those thresholds.
5–9%
the revenue increase associated with a one-star Yelp rating increase, for independent restaurants — measured against real Washington State tax revenue records, not self-reported sales.
Source: Michael Luca, "Reviews, Reputation, and Revenue: The Case of Yelp.com," Harvard Business School Working Paper No. 12-016 (2011, rev. 2016)
The effect was driven entirely by independent restaurants — chain-affiliated locations showed no statistically significant revenue change from their rating at all. Luca's own explanation: a chain already carries a standardized reputation customers can rely on for free (you know what a Big Mac tastes like before you order one), so a Yelp rating adds little new information. For an independent restaurant, the rating is often the only reputation signal a stranger has. A related paper by Luca and Georgios Zervas, "Fake It Till You Make It," found that as competitive pressure rises — especially among independents — so does the likelihood of a fake review showing up, worth keeping in mind before treating every rating jump as organic.
Do negative reviews really carry more weight than positive ones?
Judith Chevalier and Dina Mayzlin answered this with real sales data years before Luca's Yelp paper, in "The Effect of Word of Mouth on Sales: Online Book Reviews," published in the Journal of Marketing Research in 2006 (circulated earlier as an NBER working paper in 2003). They tracked sales-rank and review data for the same books sold on both Amazon.com and BarnesandNoble.com — 2,394 book-observations in total, 1,093 of which carried reviews on both sites, letting them compare how the same book performed under different review profiles on each platform.
Their own illustrative example, straight from the paper: take a book with four 5-star reviews and a sales rank of 500 at both Amazon and BN.com. Change just one of those Amazon reviews from 5 stars to 1 star, and their model predicts the Amazon rank falls to roughly 603 — a swing equivalent to about 20 fewer books sold per week at Amazon, relative to BN.com.
| Finding | What the paper actually shows |
|---|---|
| 1-star reviews | Statistically significant negative effect on relative sales at Amazon; significant at only the 6% level at BN.com |
| 5-star reviews | Coefficient "almost zero," and the wrong sign, at BN.com — essentially no measurable positive effect there |
| Why the asymmetry | The authors' own reading: positive reviews are easier to plant (by an author or publisher) and so carry less credibility than negative ones |
One version of this study that circulates online claims a one-star improvement produces "up to a 9.9% increase in sales." We read the paper itself and could not find that figure anywhere in it — not in the text, the abstract, or the results tables. It looks like a conflation with an unrelated number (possibly Luca's 5–9% Yelp figure, above). Use the rank-based example instead — it's the one that's actually in the paper.
How many stars do shoppers require now — and is the bar rising?
BrightLocal has run a Local Consumer Review Survey most years since 2017, its recent editions each fielding roughly 1,000–1,100 US consumers with a disclosed sample size — which makes it one of the few genuinely trackable, year-over-year sources on this topic, rather than a single number that gets endlessly recycled without a date attached.
The clearest, most comparable trend line in that series is the share of shoppers who won't consider a business below a 4-star rating: 57% in the 2018 edition, up from 48% the year before it. By the 2026 edition, 68% said they require at least 4 stars — and, more strikingly, the share requiring 4.5 stars or higher specifically had jumped to 31%, reportedly up from about 17% in the prior edition alone.
Read those two points as a genuine directional shift, not a smooth curve — BrightLocal's exact question wording moved between editions (a "won't use below 4 stars" cutoff isn't quite the same measurement as a "require at least 4 stars" one), and its 2024 edition, using yet another wording, found 71% wouldn't consider a business below 3 stars. The more stable, comparable figure across recent editions is review volume rather than star rating: the 2024 edition found 59% of shoppers want to see 20–99 reviews before trusting a business's average rating, and the 2026 edition found 47% won't use a business with fewer than 20 reviews — two differently-worded questions landing on roughly the same number, a better sign of a real threshold than the noisier star-rating figures. As for the ubiquitous "93% of shoppers read reviews before buying" line: it doesn't trace to one disclosed study — it's copied across vendor blogs that cite each other, with the actual number ranging anywhere from 62% to 98% depending on the source and question wording. BrightLocal's own most recent, disclosed-sample figure — 97% say they read reviews for local businesses — is the version we'd trust, and even that is self-reported behavior, not observed behavior.
Does replying to reviews actually raise your rating?
Davide Proserpio and Georgios Zervas tested this directly in "Online Reputation Management: Estimating the Impact of Management Responses on Consumer Reviews," published in Marketing Science in 2017 (a finalist that year for the John D.C. Little Award). Their dataset: the entire review history of 5,356 Texas hotels listed on TripAdvisor — 314,776 reviews in total, spanning August 2001 to December 2013 — cross-checked against the same hotels' Expedia reviews. Because TripAdvisor rolled out management responses at a fixed point in time, the timing gave the authors a natural experiment: they could compare a hotel's own ratings before and after it started replying, using a difference-in-differences design.
+0.12 stars
the average rating increase TripAdvisor hotels saw after they started responding to reviews, alongside roughly 12% more reviews being posted — in a peer-reviewed study of 314,776 reviews across 5,356 Texas hotels.
Source: Davide Proserpio & Georgios Zervas, "Online Reputation Management," Marketing Science, Vol. 36 (2017), pp. 645–665
A 0.12-star average shift sounds small until you factor in TripAdvisor's rounding: roughly a third of hotels that started responding saw their displayed, rounded rating rise by half a star or more within six months. The likely mechanism isn't that guests suddenly had better stays — it's selection: once management is visibly reading and replying, dissatisfied guests become less likely to post a short, indefensible complaint, so the negative reviews that still land tend to be fewer but longer and more substantive. The effect also showed up on Expedia's independent ratings for the same hotels, which argues against it being a TripAdvisor-only artifact. One figure we could not verify: a widely-repeated claim that review responses make customers willing to pay roughly $9 more for a comparable purchase. We couldn't trace it to a specific, citable, peer-reviewed source — treat it as unconfirmed rather than fact.
Which channel actually gets a review written — email or SMS?
The largest disclosed dataset we found on this is Birdeye's "State of Online Reviews 2025" report, built from platform data across more than 150,000 US businesses, covering requests sent between January and December 2024. It's worth being precise about what this is: real, large-sample, dated platform data from one review-management vendor — not an independent academic study, and not one with a published statistical methodology beyond its sample and date range.
Share of requests sent
Response rate
Email was still the more-used channel (60% of requests sent) but the lower responder (27%); SMS was used less (40% of requests) but converted better (38%). Two other numbers from the same report complicate a simple "just switch to text" takeaway: SMS click-through rate actually fell from 8% in 2023 to 6% in 2024 — Birdeye attributes the drop to message fatigue and rising wariness of spam and AI-generated scam texts — and businesses needed an average of 17 reminders to generate a single review in 2024, up from 16 the year before. In oToK, this argues for escalation over a single-channel bet: trigger the review request from Sale recorded (fired the moment a purchase is logged to the Sales log) or an Event attended trigger, and send it as an email with automatic WhatsApp fallback if it goes unopened — the same channel-escalation shape the response-rate gap above argues for, without committing to SMS as the default.
Is it legal to only ask happy customers, or offer a discount for a review?
This is the part of reputation marketing that trips businesses up, because the old advice used to be common: filter customers by satisfaction first, then only ask the happy ones to leave a public review. Two separate rulebooks now say no to that — one a platform policy, one actual law — and neither is obscure guidance nobody's heard of.
Google's Business Profile policy is explicit: businesses may not "offer incentives — such as payment, discounts, free goods and/or services — in exchange for posting any review," including incentives to remove or revise a negative one, and may not "discourage or prohibit negative reviews, or selectively solicit positive reviews from customers" — the practice usually called review gating. On-premises pressure to leave a review is banned too. What's explicitly allowed: asking every customer, equally, for an honest review, with no incentive attached. Violations can mean review removal, listing penalties, or a suspended Business Profile.
In the US, the FTC turned similar principles into federal law on August 14, 2024, with its "Trade Regulation Rule on the Use of Consumer Reviews and Testimonials" (16 CFR Part 465), effective October 21, 2024. It bans six specific practices: fabricated reviews (including AI-generated ones), paying for reviews conditioned on what they say, undisclosed reviews from a company's own officers or employees, a company running a fake "independent" review site, suppressing negative reviews through unfounded legal threats or intimidation, and buying fake social-media followers or engagement. Violations carry civil penalties of up to $51,744 each, as of the rule's 2024 baseline.
Israel has no direct equivalent on the books. We looked specifically for an Israeli law or enforcement action targeting fake or manipulated reviews and found none — no clause written for reviews specifically in the Consumer Protection Law, 5741-1981, and no published case from the Consumer Protection and Fair Trade Authority naming review manipulation. The general prohibition in Section 2 of that law, against any act or omission likely to mislead a consumer on a material aspect of a transaction, could plausibly be argued to cover fabricated reviews, but that's our own reading of a general clause, not a documented precedent.
Google's policy applies to any business with a Google Business Profile, anywhere — a business in Israel asking only its happiest customers for a review is exposed to Google's own enforcement (delisting, penalties) even though no Israeli statute addresses the practice directly. Treat Israel as a real regulatory gap here, not as ground you're automatically covered on. The safer default either way: one review-request flow that reaches every customer who completes a purchase or appointment, not a branch that filters by predicted satisfaction before deciding who gets asked.
How do you build a review-request flow that stays compliant?
- 1Pick one trigger that fires for every customer, not a filtered subset — order completion, a purchase logged to your Sales log, or an appointment marked attended all work; a trigger that only fires for people you already know are happy is the review-gating pattern both Google and the FTC ban.
- 2Add a delay before asking. The consensus across review-platform vendors — informed practice, not a controlled study, so treat it as a starting point rather than a proven optimum — is roughly 3–7 days after delivery for a typical purchase, shorter for a digital product, longer for a considered purchase like a service or a course.
- 3Don't commit to a single channel. Email is still the higher-volume default, but the disclosed response-rate data above favors an escalation: ask by email first, and fall back automatically to a more immediate channel if it goes unopened.
- 4Point the review link at your actual public listing — Google Business Profile, an industry review site, or wherever your customers actually check. An internal-only feedback form isn't a public review, and won't move your public rating either way.
- 5Reply to what comes back, positive or negative, on a routine cadence rather than only when a review is bad enough to worry about — the research above ties the act of replying itself, not just the tone of the reply, to a measurable ratings lift.
In oToK, a review-request flow doesn't need a dedicated feature: trigger it from Sale recorded (fired the moment a purchase is logged to the Sales log, independent of deals, payments or orders) or from an Event attended trigger, add a Wait node timed to your own delivery window, then send an email with automatic WhatsApp fallback already built in if it goes unopened — the exact channel-escalation pattern the response-rate data above argues for. Skip the Condition node here: branching the flow by a contact's tags, fields or prior sentiment before deciding who gets asked is the one pattern to avoid, since it's the shape of review gating either policy would flag.
Frequently asked questions
- Is the "5–9% more revenue per star" stat real?
- The finding is real and comes from a Harvard Business School working paper by Michael Luca that matched actual Washington State tax revenue records to Yelp ratings for around 60,000 Seattle restaurants — but it's a working paper, not a peer-reviewed journal article, and the effect only held for independent restaurants. Chain restaurants showed no statistically significant revenue change from their rating in the same study.
- Do negative reviews really outweigh positive ones?
- Yes, according to the peer-reviewed evidence we could find. Chevalier and Mayzlin's 2006 Journal of Marketing Research study of real Amazon and Barnes & Noble sales-rank data found a single 1-star review carried a measurable negative effect on relative sales, while 5-star reviews showed almost no measurable positive effect at one of the two retailers studied. We could not verify the specific "9.9% sales increase" figure that circulates online attached to this paper — it doesn't appear in the paper itself.
- Is it illegal to only ask happy customers for a review?
- It violates Google's Business Profile policy outright — review gating is explicitly banned, with review removal or profile suspension as the consequence — and in the US it can fall under the FTC's 2024 rule against review suppression and manipulation, with civil penalties up to $51,744 per violation. Israel has no direct statute on this specific practice, but Google's policy still applies to any business with a Google listing, regardless of where it's based.
- Does replying to reviews actually change your star rating?
- A peer-reviewed 2017 Marketing Science study of 314,776 TripAdvisor reviews across 5,356 Texas hotels found that starting to reply to reviews was followed by roughly a 0.12-star average rating increase and about 12% more reviews, likely because dissatisfied guests become less likely to post a quick, indefensible negative review once they know management is reading and responding. We could not verify the specific "$9 more willingness to pay" figure sometimes attached to review-response research; treat it as unconfirmed.
- What's the best channel to send a review request through?
- The largest disclosed dataset we found — Birdeye's 2025 report, built from 150,000+ US businesses on its own platform — found SMS got a 38% response rate against email's 27%, even though businesses sent more review requests by email (60% of volume) than SMS (40%). It's first-party platform data from one vendor, not an independent study, so treat the direction as informative rather than definitive — but it lines up with why an automatic email-to-WhatsApp fallback tends to outperform committing to a single channel outright.
