Methodology · Bite Score · Cardiff verdict system
How the Bite Score works.
Every published Bite Score is a weighted mean of six axes— food, value, service, portions, atmosphere and the cross-check ‘would I actually come back?’ — rescaled out of 10 and stamped with a single public verdict. GO or NO. No buyable star inflation. No half-points that pretend the experience was better than it was.
The six axes
Six things you can disagree about, then a number you can’t.
Each axis is scored in the review form on a one-to-five scale, anchored by the sentence under the title so the meaning stays the same from reviewer to reviewer. No axis scores ambiguously; no axis is editorialised.
Food
1 = inedible, leave the plate · 3 = OK but forgettable · 5 = cook I’d send a friend
Eat the dish at the table, with a fresh order — no question of context. Anchor the score to whether you would actively point someone at this menu, not to whether the photo looks good.
Value
price × portion-weight × repeat-visit signal
Not the cheapest meal in the room — the meal you'd happily pay for twice. Scaled 1–5, three sub-signals multiplied together so a £17 plate can still score 5 if the second visit pays for itself.
Service
1 = ignored, 3 = fine, 5 = remembered the order next visit
Speed AND recovery. A slow lunch where the host caught and fixed the error beats a fast lunch where the error never registered; a server who remembered an allergy next visit is an automatic 5.
Portions
1 = insult-price ratio · 3 = adequate · 5 = generous without feeling wasteful
Scored as a pair with value. A £24 tasting plate that leaves you pecking is a 1; a £14 main that fills a coat-pocket the next morning is a 5. The portion is the answer to "is this enough?"
Atmosphere
background noise + lighting + how long you comfortably stayed
Three measurements taken together. A bright, loud tasting menu still scores low if you could not wait to leave; a long quiet dinner scores high if the room released you rather than pressured you out.
Would I actually come back?
1 = flee · 3 = maybe · 5 = already planning the next visit
The cross-check. Six axes that score a 9 with a 'never again' signal still collapses down — return is the floor that catches averaged-out noise. A single binary question that anchors the verdict.
Category-specific criteria
Coffee shops
Coffee quality · Food and bakes · Service · Atmosphere · Value · Return intent
Category-specific criteria
Cafes
Food quality · Coffee and drinks · Portions and value · Service · Atmosphere · Return intent
How they collapse
Six axes become one number. One verdict.
We take a weighted meanof the six axes — food carries 1.5×, everything else carries 1× — then rescale to /10. The published verdict is a single GO or NOat the ≥ 7.0 threshold. A great meal with slow service still publishes a high score; a perfectly fine meal the reviewer will not return to reads honestly.
| Axis | Weight | Why |
|---|---|---|
| Food | 1.5× | the only axis weighted — the rest get 1× |
| Value | 1× | price × portion × repeat-visit |
| Service | 1× | speed AND recovery |
| Portions | 1× | answer to "is this enough?" |
| Atmosphere | 1× | noise + light + how long you stayed |
| Return | 1× | the binary cross-check |
The formula
(scorei× weighti) ÷ Σ weighti·× 2·
Worked example
A Cardiff cafe, Tuesday lunchtime
- Food5
- Value4
- Service3
- Portions4
- Atmosphere5
- Return5
The six axes multiply by their weight (food 5×1.5 = 7.5, the other five at ×1 for a 28.5 weighted sum), divide by the total weight of 6.5, then rescale to /10.
Published Bite Score: 8.8 / 10·
Versus a star rating
Fairer than a star rating.
A star rating collapses the meal into one number on a curve that lets one bad axis punish five great ones. Six axes let the verdict hold what was true and discard what was an outlier — and the binary GO/NO stops the review from drifting up to 4.4 on momentum alone.
Stars hide the reason.
A bad 2-star lunch could mean burnt food, hostile service or a dirty floor — all flattened into the same number. Six axes tell you exactly which part to trust and which part to discount.
A great meal survives a slow service.
Star averages punish outlier axes. The weighted mean keeps a great food score visible at a 9 even if the service that day was a 3 — the verdict still reads GO, with the reason attached.
The verdict is binary, not curved.
GO or NO. A 6.9 is the same as a 4.4 — both NO. Harder to game by eeking the average up; nothing in BwtSgor rewards a 6.5 that pretends to be a 7.
No paid placement, no edit window.
Every score is a field-tested review, no venue pays to inflate or suppress. Bite Scores change only when filed verdicts change them.