A 4.3 average across two thousand reviews tells you remarkably little. You do not know who wrote them, whether they were paid, whether they were filtered, or whether the two thousand people share anything at all with you.
A single "we went last month, the lamb was excellent, skip the dessert" from someone whose taste you know is worth more than all of it. The reason is not sentimentality. It is accountability.
What is wrong with anonymous ratings
Not that people lie — most do not. The problems are structural and they compound.
The sample is self-selecting People review when they are delighted or furious. The vast middle, which is what you would most likely experience, never writes anything.
Incentives run one way Businesses have strong reasons to solicit, filter and occasionally buy reviews. Customers have almost none to write them.
An average hides the distribution 4.3 can mean consistently good, or half excellent and half terrible. Those are completely different restaurants.
The reviewer is nobody You cannot calibrate against a stranger. Is this person hard to please, or do they rate everything five stars?
What a network adds
When a review is attached to a person you actually know, three things come with it for free: you know their taste, they know you will see it, and they will hear about it if they were wrong.
That last one is the whole mechanism. A stranger faces no cost for a careless review. Someone in your network does, and that cost is what makes the signal trustworthy.
The obvious objection
Your network is small, and it does not cover everything. You will never have a friend who has been to every restaurant in the city.
True — which is why the useful version is not "only show me friends". It is: show me what my network knows first, and be explicit about where everything else came from. The failure of the current model is not that it includes strangers. It is that it presents strangers and friends as the same thing.
Verified beats voluntary
A review from someone who definitely paid at the place is a different class of evidence from one anyone can post. Tying the review to a real transaction removes the cheapest kind of fake outright.
It does not remove every kind — but it raises the cost of faking from zero to the price of a meal, which turns out to be most of the battle.
What to look for
When you read a review, ask what the person had to spend to write it. Attention? Nothing. Their reputation with people they know? That is a review worth reading.
More Blogs
Decentralizing Influence: Empowering Everyday Users To Influence Business Decisions
April 2, 2024
Read More