The art of reviewing
A 3.7-star place with 30 reviews can be the best pizza in your neighborhood. A handful of bad reviews about parking pulled the math down. People scrolling with a 4.0+ filter never see it.
the problem with the number
Star ratings are mostly noise because of who shows up to write them. The middle gets quiet. The people who write reviews are either delighted or furious, almost never indifferent. Most ratings cluster between 3.0 and 4.5. The middle is empty because the middle didn’t bother to log in.
The other thing the score can’t carry: ratings are taste-dependent. A 5-star Italian dinner in Stockholm is a different category than a 5-star bao stand in Bangkok. The number ignores what you’re rating against.
the local-pizza problem
The most common bad review judges a place against the wrong reference. Someone reviews their neighborhood pizza spot and gives it 2 stars because the dough wasn’t Neapolitan, the wine list wasn’t curated, and the service wasn’t doting. None of that was the offer. It’s a neighborhood pizza spot. Review it as a neighborhood pizza spot.
On the pizza spot, it just signals the reviewer doesn’t know what they walked into.
what I actually do on Google Maps
A few habits that have changed my hit rate:
Sort by lowest rating first. The default “most relevant” sort hides complaints. The lowest-rating sort surfaces them. Read five or ten of them. Most of the time the complaints have nothing to do with the food: parking, weekend crowds, one bad night, the place was closed when they showed up, a host who looked at someone wrong. Those are the reviews that should not have moved your decision in the first place. If the patterns are actually about the food, take them seriously.
Be skeptical of low scores with few reviews, especially on new places. A 3.4 with 8 ratings on a newer spot can be excellent. A couple of angry reviews unrelated to the food do most of the work to pull that number down, and the math doesn’t recover until many more people show up. An older place with only 8 reviews is a different signal: low foot traffic, an older clientele who doesn’t write reviews, or something else keeping people from coming back. A 4.8 with 12 on a new place is usually a real positive signal. A 3.9 with 800 on an established place is what it looks like: OK, no better. A 3.9 on a place open two months is the same small-sample noise as the 3.2 with 8.
Look for the texture of real reviews. Bought or AI-generated reviews lean generic: “great food, will come back,” “highly recommend,” “amazing service.” Authentic reviews mention specific dishes, specific waiters, specific moments. They contradict each other. If the five-star reviews all sound the same and the one-star reviews are richer and weirder, the rating is being gamed. Another check: Google Maps shows the original language of reviews you’ve translated. If a Stockholm place has thirty glowing reviews in Indonesian or Russian, the reviews are bought.
Sort by newest if you’re undecided. Mostly to check whether the bad reviews are recent (something changed) or old (they fixed it).
what I’m trying to do when I write one
Write the review the next person actually needs. Which means: name what kind of place this is, who’d like it, what to order, what to skip. If it’s a neighborhood pizza spot, say that. If it’s a special-occasion place, say that. The star is the least informative part of the review. The sentences are the part.
I’m still bad at this. The hardest part is releasing my own context. I went to the pizza place starving on a Tuesday and loved it, or to a fine-dining place after a hard week and was distracted. Neither of those is the place’s fault. Separating the visit from the place is the part I’m still learning.