Methodology
How we score restaurants
A restaurant has no barcode and no ingredient label, so the map runs on the next best thing: structured community ratings. People who have eaten there — or asked the kitchen directly — answer the same short questions about what the restaurant cooks with, and those answers turn into a 0–100 score with published math. Chains are rated as one: a review of any location counts toward the whole chain, so a national brand builds a real score fast.
The four categories
Every rating answers up to four questions — how much of the menu uses seed oils, artificial additives, plastic contact, and premade food. “None” is a perfect answer, “Some” is halfway, “All” is the worst; for seed oils, a kitchen that swaps to better oils on request earns a bump. Each category gets its own 0–100 score, and the overall score blends them by how much they matter:
What the kitchen actually fries and cooks in. The dominant category — it moves the overall score six times as much as plastics.
Dyes, preservatives, and synthetic flavorings in the food itself.
Food prepped, stored, or served in plastic. A tie-breaker, not a driver.
Does the food arrive premade and reheated, or is it made fresh in-house? Also a tie-breaker.
Evidence beats opinion
Every answer also carries how the reviewer knows it, and better evidence literally weighs more in the math. A category’s score is the confidence-weighted average of all its answers, rescaled to 0–100 — so one public statement from the restaurant outweighs four hunches:
A gut feel from eating there. Counts, but only a quarter as much as proof.
The reviewer watched the kitchen work or read it off a label.
A cook or server answered the question directly.
The restaurant answered in writing.
The restaurant says it publicly — menu, website, or press. Full weight.
Moderated before it counts
Ratings are AI-moderated before they touch a score: only approved reviews are counted, so spam, brigading, and off-topic reviews never move a restaurant’s number. Reviewers can skip any question they are not sure about — a skipped question simply does not count, in either direction.
Unrated means unrated
A category nobody has answered yet has no score, and it is excluded from the overall blend entirely — a restaurant rated only on plastics is not “great overall”, it is simply unproven elsewhere. A restaurant with no approved ratings at all shows as Unrated on the map: gray, not a verdict. We never fill gaps with guesses.
Personalized to what you care about
Tell Tallow which of the four categories matter to you and the map re-weights every score to just those. One deliberate difference from the overall score: if you care about a category the restaurant has no ratings for, it counts as a zero for you — “unproven on the thing I care about” should drag your personal score down, not be silently ignored.
Score bands
Restaurant scores band exactly like product scores, plus the gray Unrated state for places without approved ratings yet:
Poor
0–40
Okay
41–69
Clean
70–100
No black boxes, no paid placements. We will never accept payment from the brands or restaurants we rate.