For twenty years, every business chased the same thing: a perfect five-star rating. It turns out shoppers stopped believing it. Northwestern’s researchers found that the likelihood someone actually buys peaks below five stars — because, somewhere along the line, we all learned that perfect doesn’t mean good. Perfect means somebody’s hiding something.
In this episode of Marginally Better, Joe Taylor, Jr. unpacks the great review reversal: the research showing why a few imperfect reviews sell more than a flawless score, the Fashion Nova cover-up that ended in a $4.2 million settlement — the FTC’s first-ever case for hiding negative reviews — and why the three-star reviews most businesses dread are the most useful feedback they’ll ever get for free.
Episode Links:
The Research: Why Perfect Doesn’t Sell
- Northwestern University Medill Spiegel Research Center — How Online Reviews Influence Sales
- PowerReviews & Northwestern’s Spiegel Research Center — Five-Star Product Ratings Are “Too Good to Be True”
- PowerReviews — How Fake Reviews Destroy Consumer Trust
The Trust Crisis
- Pangram Labs — Three Percent of Front-Page Amazon Reviews Are AI-Generated
- Bazaarvoice — Shopper Experience Index 2025
The Law
The Fashion Nova Case
- FTC — Fashion Nova Will Pay $4.2 Million to Settle Allegations It Blocked Negative Reviews (January 2022)
- FTC — Finalizes Order with Fashion Nova Over Blocked Reviews (March 2022)
- FTC — Sends Refunds to Consumers Affected by Fashion Nova’s Deceptive Review Practices (January 2025)
The Amazon Origin Story
Products Mentioned
Transcript:
[00:00:00] For twenty years, every business on earth has been chasing the same thing. Five stars. The perfect score. That flawless little row of gold that’s supposed to tell a customer: trust us, everybody loves us, nothing ever goes wrong here. And it turns out the customer doesn’t believe a word of it.
Researchers at Northwestern have been watching how people actually shop, and they found something that should make a lot of marketing departments sit down. The rating that makes someone most likely to buy isn’t five stars. It’s lower. There’s a sweet spot, and once you cross it, heading up toward perfect, people start to back away. Because somewhere along the line, we all learned that perfect doesn’t mean good. Perfect means somebody’s hiding something.
So today, the great review reversal. I’ll tell you about the company that paid four point two million dollars to bury the very thing that would have sold more clothes. I’ll show you the number Northwestern put on perfection. And I’ll make the case that the worst reviews you’ve got might be the most valuable thing you own. That’s all coming up on Marginally Better.
We have built an entire economy on the star rating. We check it before we book a hotel, buy a drill, pick a taco place, hire a plumber. The little number does an enormous amount of work. It stands in for the trust we used to get from a neighbor or a salesperson we knew. And because it carries that much weight, a lot of companies decided the smart move was to make their number as close to perfect as possible, by whatever means necessary. That instinct is about to get a lot of them in trouble.
So start with the research, because it flips the game. Northwestern University’s Spiegel Research Center studied how online ratings actually move buyers, drawing on a huge pool of real review data. And they found that the likelihood someone buys a product peaks when its average rating sits between about four and four point seven stars. Not five. After that peak, as the score climbs toward a perfect five, purchases actually start to slide back down.
Their phrase for it was blunt. A five-star rating is, to the shopper, too good to be true. The presence of a few less-than-glowing reviews doesn’t scare people off. It reassures them. So the thing companies have been chasing the hardest may be the thing quietly costing them sales. Which raises a question worth sitting with. If perfect isn’t what convinces people, what exactly have we all been optimizing for?
And this isn’t some subtle effect only an economist can see. Shoppers will tell you outright. PowerReviews surveyed buyers about the ideal rating, the score that would actually make them comfortable hitting buy. Just six percent said a perfect five out of five. Six. The biggest group, forty-four percent, said the sweet spot was somewhere between four and a half and just under five. Another forty-two percent put it lower still, between four and four and a half.
Add it up. More than eight in ten shoppers are telling you, in plain language, that they want to see a flaw or two before they trust you. The customer isn’t looking for a company that never disappoints anyone. They’ve lived long enough to know that company doesn’t exist. They’re looking for a company that’s honest about the disappointing.
Part of why people got this skeptical is that they’ve been burned. The reviews themselves are increasingly suspect. Pangram Labs, a firm that builds tools to detect machine-written text, scanned thirty thousand reviews across Amazon’s best-sellers and estimated about three percent were written by AI. Now, that’s a detection estimate, not a confession, so hold it a little loosely. But one finding fits everything else today. Those AI-written reviews were far more likely to hand out a perfect five stars. About three-quarters of them did, versus roughly six in ten of the human reviews. The fakes gravitate to perfect. The people don’t.
And shoppers can feel it. Bazaarvoice, which runs reviews for thousands of major retailers, found in its 2025 shopper survey that the single most common frustration in the entire buying journey, ahead of price, ahead of shipping, was just trying to figure out whether the reviews are real. The tool we built to create trust has become the thing people trust the least. So the question stops being how do we get more five-star reviews, and becomes something harder. How do we prove the ones we have are honest?
And now there’s a law with teeth. In 2024, the Federal Trade Commission finalized a rule that flat-out bans fake reviews. Made-up testimonials, bought reviews, AI-written praise, and suppressing real ones. And for the first time, it gave the agency the power to come after violators with civil penalties. By late 2025, the FTC was already sending warning letters and naming names, including a case against a company that sold an AI tool built to crank out reviews on demand.
So put it all together. The research says perfect doesn’t sell. The customers say perfect makes them suspicious. The fakes have poisoned the well. And the government has started fining people for gaming the system. Every signal points the same direction, toward honesty. Which makes the story I want to tell you next so striking. Because it’s about a company that, right as all of this was becoming clear, spent four years and a fortune doing the exact opposite.
When we come back, I want to take you inside a wildly successful fashion brand that built a secret filter for one purpose: to make sure you only ever saw the happy customers. It worked perfectly, until the day the Federal Trade Commission knocked, and it became the first case of its kind in American history. That’s after the break on Marginally Better.
Quick word about something we built over at Johns and Taylor, because it sits right on top of what we’re talking about today. Everything in that last segment comes down to one word. Trust. And here’s the uncomfortable part. You can’t add trust to your business at the end, like a coat of paint. It has to be built into the thing itself. The way you handle a complaint. The way your site answers the question a customer’s actually worried about. Whether the experience tells the truth about what you’re like to deal with. We call that trust architecture, and most businesses are building it by accident.
That’s why we built the Experience Helpdesk. It isn’t a course you’ll never finish, and it isn’t a big agency retainer. It’s expert UX help, on demand. You send a question when it comes up, a quick voice note, a short video, or just text, and you get a real answer back, usually within one business day. The should-we-even-show-our-negative-reviews questions. The is-this-developer-quote-reasonable questions. The I’m-stuck-between-two-designs-and-I-need-a-tiebreaker questions. The ones a course can’t answer and a meeting is too slow for. Members also get our full resource library, more than thirty guided playbooks across nine areas of customer experience, plus a five-minute challenge every Monday to keep momentum going. We’re opening it with a charter rate for the first members in the door, and that rate is locked for as long as you stay. If you’ve ever needed a second set of expert eyes and had nowhere to send the question, this is for you. Come see it at experiencehelpdesk.com. That’s experiencehelpdesk.com.
It’s Marginally Better. I’m Joe Taylor Jr.
Let’s go back to around 2015, to one of the great business rockets of the last decade. Fashion Nova. If you don’t know it, picture this. A Los Angeles clothing company that figured out Instagram before almost anyone else in fashion. They’d see a look on a celebrity one night, have an affordable version on their site within days, and put it on hundreds of influencers and everyday customers who posted themselves wearing it. The founder, Richard Saghian, built one of the most-searched fashion brands on the planet, largely out of a feed full of real people in real outfits. It was fast, it was cheap, and it was enormous.
And a business like that lives and dies on one thing. The look of the thing. The vibe. The sense, when you land on a product page, that everyone who bought this dress was thrilled with it. Their entire engine ran on aspiration. You should want to be the person in this photo.
So now imagine you’re running that company, and the reviews start coming in. Most are good. But some aren’t. The fabric’s thinner than it looked. The fit runs small. The color’s off. Ordinary, honest, three-star reality, the kind every clothing company on earth collects by the truckload. And it’s landing right in the middle of your gleaming, aspirational storefront. So, a fork in the road. You can let it show. Or you can make it disappear. Fashion Nova made a choice. And the tool to carry it out was already sitting in the shopping cart, so to speak.
The FTC later laid out how it worked, and the design is almost elegant in how plainly dishonest it was. Fashion Nova used a third-party review system to manage what showed up on its product pages. And they set it so that four-star and five-star reviews posted automatically, straight to the site, no human needed. But anything lower? Three stars, two stars, one star? Those got held back. Routed into a queue. Held for approval.
And then, according to the government, nobody approved them. From late 2015 until November 2019, four years, hundreds of thousands of lower-rated reviews simply sat there. Written by real customers, about real products, and shown to no one. The storefront stayed sunny. Every page looked like everybody loved everything. By design.
And now the part that turns this from a compliance story into something almost tragic. While Fashion Nova was burying its imperfect reviews, the research was landing in public. By 2017, right in the middle of that four-year blackout, Northwestern’s Spiegel center had published the opposite of what Fashion Nova believed. That perfect ratings read as fake. That buying peaks below five stars. That a few critical reviews actually build trust. The science was on the table. Fashion Nova was spending real time and real money to manufacture exactly the flawless, suspicious storefront that researchers had already shown sells less.
They weren’t just being dishonest. They were being dishonest in the wrong direction. They paid to suppress the asset. Those three-star reviews, the fit runs small, the fabric’s thin, that’s not damage. To a shopper, that’s the most useful information on the page. It’s the stuff that tells you to size up before you buy. Hide it, and you don’t look perfect. You look like you’re not telling the truth, which, as it happens, is exactly what was going on. And a thing that big, hiding that much, for that long, eventually somebody notices.
In January of 2022, the Federal Trade Commission brought down the hammer. Fashion Nova agreed to pay four point two million dollars to settle the charges. And the detail that matters most, the one that puts this episode in the history books: the FTC said this was its first-ever case against a company for hiding negative reviews. First one. Ever. After two decades of online reviews, Fashion Nova earned the distinction of being the cautionary tale, the example the government chose to make.
The settlement didn’t just take the money. It changed the rules of their storefront. Fashion Nova was ordered to post its reviews, the critical ones included, the only exceptions being things like obscene or off-topic content. The filter came off. The three-star reality their customers had been writing for years finally had to show up on the page.
And the math should haunt anyone tempted to do the same thing. They spent four years building and running a system to hide honest feedback. They paid four point two million dollars when it came apart, and the FTC sent roughly two point four million of that straight back to harmed customers, refund checks going out as recently as 2025. And the entire time, the thing they were hiding was the thing that would have made the page more believable and, by the research, more likely to sell. They paid a fortune to make their business worse.
Saghian’s company survived it. They’re still big. This isn’t the end of Fashion Nova. But it’s a permanent line in their story, and it’s a gift to the rest of us, because it makes the lesson impossible to miss. You cannot buy trust by deleting doubt. The moment you scrub away every imperfection, you haven’t removed the customer’s hesitation. You’ve just removed their reason to believe you. So the question every business should ask itself, before it reaches for the filter, is the one Fashion Nova answered wrong for four straight years. When a customer tells the truth about you, in public, is your first instinct to listen, or to hide it?
I want to leave you somewhere more hopeful than a courtroom, because the flip side of this is good news, the real kind. The most valuable feedback your business will ever get is sitting in your worst reviews, and it’s free.
There’s a piece of internet history worth knowing here. Back in 1995, when Amazon was a tiny operation selling books out of a converted house, Jeff Bezos made a decision that horrified the publishing industry. He let customers post negative reviews. Right there on the page, next to the buy button. Someone wrote him an angry letter, the gist of which was: you don’t understand your own business, you make money when you sell things, so why are you letting people trash them? And Bezos’s answer, as he later recounted it, became one of the most radical ideas in modern retail. We don’t make money when we sell things, he said. We make money when we help customers make purchase decisions.
Sit with that. He didn’t see the negative review as a threat to the sale. He saw it as the service. The bad review is how the customer trusts the good ones. Strip out the critical voices and you don’t have a more persuasive store, you have a less believable one. Thirty years later, the rest of the economy is finally catching up to what he understood in a rented house.
So here’s something you can actually use this week, and it costs nothing but a little courage. Go read your own three-star reviews. Not the five-stars, those just feel nice. And not the one-stars, half of those are someone having a bad day. The three-stars. The people who basically liked you but hit something that bugged them. That’s the most honest signal you will ever get about your own business, written by people with no reason to flatter you and no axe to grind. They are telling you, for free, exactly what to fix.
And when you find a fair criticism out in public, don’t reach for the filter. Answer it. A thoughtful reply under a critical review does more for the next shopper than a hundred glowing ones, because it shows them what you’re actually like when something goes wrong. Which is the only thing any of us really wants to know before we hand over our money.
That’s Marginally Better for this week. If you take one thing with you, take this. Trust was never the absence of flaws. It’s what you do with them out in the open. The perfect storefront isn’t the convincing one. The honest one is. And building that kind of honesty into your business on purpose, into your site, your reviews, the way you handle the moment something goes wrong, that’s hard to do from the inside, because you’re too close to see it. If today got you wanting a second set of expert eyes on yours, that’s exactly what we built the Experience Helpdesk for. You know where to find us.
Thanks for listening to Marginally Better. If you liked what you heard, do us a favor and leave a review on Apple Podcasts, and be honest. A four-star with a note about what we could do better helps us more than a perfect score ever could. We practice what we preach around here. For behind-the-scenes notes from me and the team, head to marginallybettershow.com, or just follow the link in the show notes.
Marginally Better is a Calufrax Radio production. Our producer is Nicole Hubbard, with research by Connie Evans. I’m Joe Taylor Jr.
