Star rating30pts
The business's public Google average, scaled directly.
Methodology
One published formula, applied identically to every business in every category and every city on this site. It is never adjusted by hand for a specific business. When the formula changes, the version number changes with it, and every ranking is recomputed under the new version, not patched.
The six signals
Five signals add points, weighted out of 100. One signal subtracts, and decays to nothing over time rather than following a business forever.
Star rating30pts
The business's public Google average, scaled directly.
Rating confidence25pts
How many reviews back that average up, on a log scale. This is the signal the worked example above is measuring.
Review volume20pts
Total review count on its own scale, separate from confidence, so a business that clearly dominates its category on review count keeps earning credit for it.
Recent growth15pts
New reviews per day since the last refresh. A business with no prior snapshot yet scores this signal at half credit, neither punished nor rewarded for a history it doesn't have.
Verified operator10pts
A flat bonus for confirming you actually run the business. Never inferred from Google data, only from our own claim process.
Recent negative reviewssubtracted
A penalty built from low ratings in the small, recent sample Google's API returns, decaying to zero after 180 days. Explained in full below, it is deliberately not treated as a fixed weight because it is a proxy, not a count.
Scope
The six signals above are the whole of what counts. These are the two things worth stating that are not in them.
Google's Places API returns an average rating and a total review count, but no breakdown of how many 1-star or 5-star reviews a business has, and at most 5 individual reviews per business, not its full history. Our low-review penalty is built from that small visible sample, not a true count of every negative review, and we can't reliably detect purchased or incentivized reviews. We say so here rather than imply a precision we don't have.
Questions
The person who publishes these rankings does also sell marketing help to businesses, on this site's for businesses page. Two things make that checkable rather than something you have to take on trust. The formula is published above in full, including every weight, so any ranking here can be recalculated from a business's public Google data by someone who does not believe us. And the only inputs are that public data: there is no field anywhere in the system for "is a client". We do not sell rank, we sell help earning the reviews rank is built from. If you spot a ranking that does not follow from the formula, that is a bug or a lie, and either one is worth telling us about.
Only that whoever claimed the listing controls a phone number or WhatsApp account matching the business, proof of operator, not proof of qualification. A credential check against the relevant regulator is planned as a later, separate badge, and will be clearly distinguished from this one when it ships.
Four reasons, all of them ours rather than yours. We find businesses by searching Google across a grid of points covering the city, so a listing Google does not return for that category is one we never see. We drop a result whose Google primary type is not the category, which is what keeps a gym out of a physiotherapy list even when Google tags it with both. We drop a business whose Google address does not name a municipality in the city, because we will not claim it is somewhere we cannot show it is. And a listing with very few reviews scores near zero, so it is ranked, just far down. If you think you should be here and are not, that is worth telling us.
A refresh re-reads every business in a category from Google and rescores it. Weekly is the intended cadence, chosen deliberately over daily because review counts do not move fast enough for daily to mean anything. Every ranking page shows the date of the refresh behind it, so you never have to take the cadence on trust — read the date.