An NT$80,000 Lesson
A restaurant owner spent NT$80,000 to hire a marketing agency that posted 200 five-star reviews. His Google rating jumped from 3.8 to 4.6.
The reservation phone wouldn’t stop ringing.
Three months later, the rating dropped to 3.2 — lower than before he started. Every new negative review was harsher than the ones before: “Lured here by the good reviews, but it’s nothing like what they claim.” “Extremely disappointed.” “I’ll never trust ratings again.”
What NT$80,000 bought wasn’t good reviews. It was a collapse of trust.
Worse, during those three months, his Google Business Profile accumulated a large number of “fake review” reports. Google’s anti-fraud algorithm kicked in — of the 200 five-star reviews, 130 were flagged as suspicious (from new accounts, no other review history, IP addresses concentrated in the same range). These reviews were eventually deleted in bulk by Google. But by the time they were removed, his rating had already plummeted to 3.2, because the real negative reviews had already overwhelmed everything else.
He didn’t just lose the NT$80,000 marketing fee. He lost three months of operating time — three months in which he attracted the wrong customers, generated angry negative reviews, and then got “corrected” by the algorithm to a position worse than where he started. If he had spent that NT$80,000 during those three months on improving service speed, training staff, or optimizing the menu, his 3.8 rating might have naturally climbed to 4.0.
But he chose the shortcut. The shortcut took him in a full circle, back to a place worse than where he began.
The Death Spiral of Fake Five-Stars
Let’s break down the NT$80,000 journey into a timeline, and you’ll see a clear downward spiral structure.
Week One: 200 five-star reviews go live. Rating jumps from 3.8 to 4.6. Search ranking rises, visibility increases. The reservation phone starts ringing. The owner feels the NT$80,000 was well spent.
Weeks Two to Four: New customers flood in. But their expectations are set at a 4.6 level. The actual experience is a 3.5 to 3.8 standard. A gap emerges. The first wave of negative reviews appears — and they’re more intense than usual, because customers feel “deceived.”
Month Two: Negative reviews accumulate. Rating begins to slide: 4.6 → 4.4 → 4.1. Meanwhile, Google’s anti-fraud system starts analyzing the sources of those 200 five-star reviews. New accounts, no review history, IP concentration — each one is a red flag.
Month Three: Two things happen simultaneously. Google deletes 130 suspicious reviews in bulk. Real negative reviews continue to accumulate. The two forces combine, and the rating plummets to 3.2.
3.8 → 4.6 → 3.2.
This isn’t a linear decline. It’s a “false rise, then retaliatory fall” curve. The rise caused by fake reviews is hollow, but the fall caused by real negative reviews is solid. The height you borrowed with fake numbers, you ultimately repay with real trust. And the interest rate is steep.
Repairing Costs Ten Times More Than Breaking
How long did it take that restaurant owner to recover his rating from 3.2 back to 3.8?
Six months. And the process was far more painful than he imagined.
Because 3.2 is a toxic starting point. Potential customers see 3.2 and their instinctive reaction is “something’s wrong with this place” — they won’t even give it a try. Foot traffic drops, revenue drops, and the budget available for quality improvements shrinks. This is a vicious cycle: low score → fewer customers → less revenue → weaker ability to improve → lower score.
To break this cycle, he did three things. First, he retrained the service staff — cutting average ticket time from 25 minutes to under 15 minutes. Second, he proactively replied to every negative review, acknowledging the problem, explaining the corrective measures, and inviting the customer back. Third, he analyzed review trends weekly to confirm whether the improvements were actually reflected in the rating.
After six months, the rating returned to 3.6. It took another three months to get back to 3.8.
From 3.2 to 3.8, he spent nine months. Going from 3.8 to 4.6 with fake reviews took only three days.
Breaking is fast. Repairing is slow. That’s the hidden cost of fake reviews — they don’t just deceive customers, they overdraft your future time and resources for repair.
The Higher the Expectation, the Bigger the Gap
The more vivid your fake reviews — “beautiful ambiance, attentive service, exquisite cuisine” — the higher the expectations of the customers you attract.
Higher expectations mean a bigger gap with the experience.
Richard Oliver’s Expectation Disconfirmation model, proposed in 1980, explains this: consumer satisfaction isn’t determined by the absolute experience, but by “experience minus expectation.”
- Experience > Expectation → Positive surprise → Satisfaction surges → Five-star review
- Experience = Expectation → Meets expectations → Average → Three or four stars
- Experience < Expectation → Negative gap → Dissatisfaction surges → One-star review
Those fake five-star reviews you bought are, in essence, systematically inflating the expectation of every new customer. Which means amplifying the force of every negative review.
A customer who would have given three stars gets pulled to a five-star expectation by your fake reviews, but only receives a three-star experience — what they give isn’t three stars, it’s one star.
There’s a key mathematical relationship here. Assume your restaurant’s objective experience level is fixed (say, “mid-to-upper range”). When you raise expectations from 3.8 to 4.6, the “expectation minus experience” gap widens by 0.8. In Oliver’s model, the larger the negative gap, the more satisfaction doesn’t decline linearly — it accelerates downward. A customer who expects 4.6 but only gets a 3.5 experience will be far more dissatisfied than one who expects 3.8 and gets 3.5. The former gives one star, the latter gives three. Same restaurant, same experience, but a two-star difference in rating purely because of different expectations.
Right Fit
It’s impossible to attract an all-encompassing customer base.
A budget stir-fry restaurant shouldn’t attract customers expecting Michelin-level service. A boutique hotel shouldn’t attract travelers expecting hostel prices.
Your real rating is a natural filter. A 3.8-rated restaurant naturally attracts customers who “know what they’ll get for NT$300.” They come, feel it’s worth it, give four stars, and come back next time.
But when you inflate the rating to 4.6, you attract customers who “expect NT$800 quality for NT$300.” They come, feel it’s not worth it, give one star, and never return.
The worst thing you can do is win customers who don’t belong to you and lose the customers who do.
And this damage is ongoing. After those “deceived” customers leave one-star reviews, your rating drops to 3.2. Now even your original target audience — the people “willing to spend NT$300 on a decent meal” — are scared off by the 3.2. They won’t read the details of the negative reviews. They just look at the number. To them, 3.2 means “probably has issues.”
Fake reviews don’t just attract the wrong customers. They also pollute the judgment of the customers you could have served well.
What Do You Actually Get from Incentivized Reviews?
Another common practice is much milder than hiring a marketing agency — offering freebies in exchange for positive reviews. “Check in with a five-star review for a free dessert.” “Show a screenshot of your five-star Google review for 10% off.” Many owners think there’s nothing wrong with this, and even consider it “encouraging customer feedback.”
But the effect is worse than you think.
What customers leave isn’t their genuine feeling. It’s “positive text written for the sake of getting dessert.” These reviews share several characteristics: short length (usually under 30 words), lacking specific details (no specific dish names, no descriptions of service scenarios), and templated language (“delicious,” “recommended,” “nice environment”). In the eyes of potential customers — an increasing number of whom have developed detection skills — these are instantly recognizable as “conditional positive reviews.”
The more serious problem: reviews filtered through “freebie conditions” systematically exclude dissatisfied customers. A customer who found the experience mediocre might still write a five-star review for the dessert. But a customer who had a terrible experience takes the dessert but can’t bring themselves to write five stars — they choose not to write at all.
So the rating you ultimately get is a “dessert-distorted score.” It’s higher than the real rating, but every review in it is less credible than a genuine one. And Google’s algorithm is also learning to identify these “incentivized reviews” — review times concentrated in specific periods, accounts in the same geographic area, high similarity in review text — these are all signals.
The cost of incentivized reviews isn’t the price of that plate of dessert. It’s the long-term health of your review ecosystem.
How Google Catches You Buying Fake Reviews
You might wonder: will Google really notice? It’s only 200 reviews, not 20,000.
Yes. And faster than you think.
Google’s anti-fraud system doesn’t rely on manual review. It uses machine learning models that analyze dozens of features for every review:
Account features — Is the reviewer a new account? Has this account left reviews elsewhere? How many? If an account has only reviewed your one establishment, that’s a red flag. Normal reviewers leave reviews in multiple places — restaurants, clinics, supermarkets. An account that “only reviewed you” is indistinguishable in the data from an account “hired to review you.”
Temporal features — How quickly did 200 reviews arrive? If they all went live within three days, when you normally get 5-8 new reviews per week, that 40x spike is an obvious anomaly signal.
Geographic features — Where are these reviewers’ IP addresses concentrated? Marketing agencies typically operate from the same office, with IPs concentrated in the same range. Real customers come from all directions — different cities, different ISPs.
Language features — Are the word patterns of the reviews similar? Marketing agencies typically use the same set of templates with slight rewrites: “Beautiful ambiance, attentive service, exquisite cuisine.” “Great value, will visit again.” “The owner puts heart into it, highly recommended.” Machine learning can detect this linguistic similarity — even if each review looks different on the surface, their syntactic structures and vocabulary distributions will be highly consistent.
In 2023, Google announced on their official blog: they block more than 200 million policy-violating reviews every day. Their anti-fraud team has over 100 people, supported by AI models scanning 24/7. The marketing agency you hired for NT$80,000 is up against Google’s hundred-person team plus AI. The odds aren’t good.
The Expectation Disconfirmation Model
Richard Oliver, 1980, published in Journal of Marketing Research, Volume 17, titled “A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions.” This model later became the foundational framework for service industry satisfaction research, cited over 10,000 times, making it one of the most cited papers in marketing history.
Core formula: Satisfaction = Perceived Experience − Expectation
This means you can improve satisfaction in two ways: improve the experience, or lower expectations.
Most people only think of the first. But the second — managing expectations — is often lower-cost and more effective.
Oliver later noted in his 1997 book Satisfaction: A Behavioral Perspective on the Consumer that expectations come from three sources: past experience, word of mouth, and brand promises. Those fake five-star reviews you bought contaminate the “word of mouth” component. When word of mouth is injected with false positive signals, expectations are systematically inflated. And the higher the expectation, the more painful the fall.
A three-star hotel that clearly tells guests “We’re a three-star property, not five-star. The rooms are small, but our breakfast is the best in the neighborhood” — guest expectations are calibrated. They arrive, find the room is indeed small (meets expectations), but the breakfast is genuinely excellent (exceeds expectations). Satisfaction is actually higher than those lured in by fake five-star reviews.
This is why Marriott’s Moxy Hotels say directly in their marketing “Small rooms, big experience” — they proactively manage expectations, turning “small rooms” from a disadvantage into a positioning signal that says “what you should anticipate are social spaces and design.” The result: guests don’t complain about small rooms because they already knew. Instead, they’re pleasantly surprised by the lobby bar and design details.
You Don’t Need to Follow What Competitors Are Doing
Your competitor might be buying fake reviews. You see it, you get anxious — their rating is 0.5 higher than yours, customers might go to them first.
But let’s think about it from another angle.
Their fake high score attracts customers with distorted expectations. These customers will leave them harsher negative reviews. In three to six months, their rating will experience a collapse — lower than before they started buying reviews. At that point, your real rating becomes the most credible number in the market.
This isn’t hypothetical. It’s the endpoint almost every business that buys reviews arrives at. The only difference is timing — depending on how long it takes Google’s algorithm to catch them.
You don’t need to jump off the cliff with them. What you need to do is: when they collapse, your real rating is stable, credible, and trusted by customers.
Trust compounds. It accumulates slowly, but once built, it’s more resilient than any fake score.
The Only Thing You Can Rely On Long-Term
Don’t be afraid of your real rating.
A 3.8-rated restaurant in a 3.8 market attracts 3.8-level customers. They come and feel it’s worth it — because expectations are calibrated.
A 4.2-rated restaurant that inflates to 4.6 through fake reviews attracts 4.6-level customers. They come and feel deceived — because expectations are distorted.
Your real rating isn’t your weakness. It’s your positioning tool. Use it to attract the right people, not to deceive the wrong ones.
This isn’t a moral issue. It’s business logic. Google’s algorithms are getting better at detecting fake reviews — account age, review frequency, IP distribution, language patterns — these are all detection signals. In 2023, Google announced they block over 200 million policy-violating reviews per day. Those five-star reviews you bought will eventually be deleted. And the moment they’re deleted, your rating will experience a “cliff drop” — because when fake reviews are cleared, the proportion of real negative reviews suddenly surges.
The only thing you can rely on long-term is real reviews left by real customers. Those reviews might not be perfect, and some might be hard to read. But they’ll help you filter for the “customers who are right for you.” The right customers come back. And customers who come back are the foundation of your revenue.
If You Can’t Buy Reviews, How Do You Improve Your Score?
You know you can’t fake it. But your score genuinely needs to go up. What do you do?
The answer is: you must first know where your negative reviews are concentrated.
Out of 100 negative reviews, 80 might stem from 2-3 structural problems. If you don’t know which 2-3, you’ll be like a headless fly — changing the menu today, switching servers tomorrow, redecorating the day after — doing a little of everything, doing nothing thoroughly.
Categorize your negative reviews. Find the root causes. Act on the root causes.
A hot spring resort did exactly this. They categorized 60 negative reviews and found 63% were related to “service attitude.” Using the Five Whys method, they traced the root cause to “insufficient staffing on holidays leading to front desk overload.” They brought back part-time holiday staff — spending an extra NT$30,000 per month, but “poor attitude” complaints dropped 70% within two months. The rating went from 2.8 to 4.1.
Not through fake reviews. Through finding the root cause and treating it directly. The cost was lower than NT$80,000, and the effect lasted longer than buying reviews.
This path is slower. But it doesn’t backfire.
Leave Your Email to Test Your Review Authenticity
Is your high rating real? Leave your email to unlock a six-question self-assessment plus an expectation calibration worksheet, and see where your review health stands — and how far your marketing promises are from the actual experience.
Further Reading
← A beauty salon’s five-star reviews might be hiding your most dangerous weakness
Fake five-star reviews are dangerous, but real five-star reviews can also mask problems. Learn how dimensional analysis reveals the weaknesses behind five stars.
→ Read
Loading…