Skip to content

THE METHODOLOGY · ISSUE №01 · FIELD GUIDE

How Harry scores.

No 'AI'. No 'algorithm'. Just opinion, with some maths to back it up — the whole game, workings out shown.

From the editor's desk

Section I

Two scores, two questions.

The quality question

Harry Hungry Score

Is it any good?

Range: 0 → 100


A weighted blend of every voice that's already written about the place. Each source pulls its own weight — literally — and Harry adjusts for noise, sample size, and the louder forms of self-promotion.

Sources are normalised to a 0–100 scale, weighted, blended, then nudged up or down by the anomaly flags in Section III.


The mix

Google ratingThe civic vote. Volume — when it shows up.
35%
Yelp ratingSharper, crankier, occasionally unhinged. Useful for moderation.
20%
TripadvisorTrusted lightly. Skewed by visitors with the wrong shoes.
10%
Editorial criticsEater, Infatuation, the broadsheets, the Michelin lot.
25%
Reservation & recent buzzHard-to-book, trending up, written about this week.
10%

The value question

Worth It Score

Is it worth what it costs?

Range: 0 → 100


Price isn't penalised on its own — Harry doesn't think dinner should be cheap, he thinks it should be fair. Four signals decide the Worth It Score, and the badge below tells you what kind of value you're getting.

A £180 omakase can be Worth the Splurge. A £14 burger can be False Economy. The price tier is context, not the verdict.


The mix

SatisfactionQuality of cooking, consistency, service, ambience.
55%
Fairness of the billPortions, comparable nearby options, whether you'd feel mugged.
20%
DistinctivenessHard to replicate — the thing the chain down the road can't do.
15%
Hype taxThe premium you're paying for the photograph, not the dinner.
10%

Section II

The blend.

Per kilo of finished score, this is the kitchen.

─ HARRY HUNGRY SCORE — THE RECIPE ─

Yield: one rating, 0–100. Serves: a hungry party of one.

350gGoogle rating, lightly skimmed
200gYelp rating, salt-corrected
100gTripadvisor, sieved twice
250gEditorial critics, picked through
100gRecent review sentiment, fresh

─ Adjustments to taste ─

  • +A pinch of distinctive cooking
  • A fistful of tourist-trap signals
  • The entire kitchen if it's spotty
  • Generous tablespoon of hype, where it's earned

Method: weigh, blend, rest, taste. If something's missing — no Tripadvisor presence, no critics — Harry redistributes the weight across whatever's left and lowers the confidence accordingly.

filed: kitchen no. 01

Section III

When the sources disagree.

Eight specific arguments Harry has with the data, every time he builds a list.

  1. Flag: google-high-yelp-low

    Google high, Yelp low

    Google says four-point-six. Yelp says three-point-two. This is usually a Bay Area thing — Google reviewers gave it five stars because they had a fine evening. Yelp users gave it one star because the host said 'good evening' to the wrong table. Harry trusts the delta itself as a signal: when it's big, expect a place that's polished on the surface and uneven once you sit down.

    Harry knocks 3 points off until the kitchen settles down.

  2. Flag: yelp-high-google-low

    Yelp high, Google low

    The reverse. Yelp loves it, Google's cool. Often a neighbourhood favourite that hasn't been found by the lunch-crowd-with-strollers yet, or a place whose regulars have an axe to grind in the comments. Harry leans toward the locals here, but he's not naive about it.

    Adjustment: −3. Same delta, different direction, same caution.

  3. Flag: tripadvisor-high-locals-low

    Tripadvisor high, locals low

    The classic tourist trap. Tripadvisor's giving it five stars and a chef's-kiss banner, the actual people who live in the postcode are quietly going somewhere else. This is the trattoria in Trastevere with the laminated menus and the man outside saying 'come in, come in, my friend'. Run.

    Adjustment: −2. Possibly more if the laminated menu has photographs.

  4. Flag: critic-loves-crowd-divided

    Critic loves it, crowd's divided

    Eater gave it a rave. The room is split. Usually one of two things: a chef doing something genuinely interesting that not everyone clocks, or a chef doing something interesting that doesn't actually taste very good. Harry will tell you which side he's on, but you should know going in that the table next to you might disagree loudly.

    Adjustment: −1. Small, because sometimes the crowd is wrong.

  5. Flag: low-sample-size

    Not enough people have eaten here

    Fewer than 25He needs about 250 reviews before he'll quote it with a straight face. Below the floor he'll still recommend the place if the editorial pile is strong, but the number on the badge gets a polite asterisk.

    Adjustment: −4. The single biggest auto-deduction in the engine.

  6. Flag: rating-recently-trending-down

    Rating's been falling, lately

    The all-time average is fine. The last six months are not. A chef left. A landlord raised the rent. The owner's son took over. Whatever it is, the people writing reviews this quarter are quieter, or angrier, than the average. Harry takes the all-time number with a pinch of salt.

    Adjustment: −3. Could be a blip; could be a slow-motion closure.

  7. Flag: tourist-skew

    Loved by tourists, ignored by locals

    Heavy visitor footfall, glowing out-of-town reviews, suspiciously few comments from people who say 'my local'. Often perfectly fine — sometimes a museum-grade clip joint. Harry'll still list it if the food's good, but he'll tell you you're paying the view tax.

    Adjustment: −2. Pair this with a Tourist Tax Warning badge below.

  8. Flag: hype-risk

    Hype risk

    Queue down the street. A waiting list of six weeks. Everyone you know has posted the same dish. The food might be excellent — but the line between 'genuinely brilliant' and 'genuinely brilliant marketing' is thinner than the menu suggests. Harry's not anti-hype, he's just allergic to paying for it twice.

    Adjustment: −2. Heavier if the queue is part of the photograph.

Section IV

Worth It, badge by badge.

Eight verdicts on the bill. The number gets you within ten points; the badge tells you what kind of value it is.

  • Great Value

    Punches well above its bill. You'd happily pay more.

    For instance:The Vietnamese on the corner doing a £9 bún chả that ruins every other lunch in the postcode.

  • Smart Spend

    Fair price, fair plate, no surprises either way.

    For instance:A neighbourhood Italian where the carbonara is correct, the wine is fine, and the bill is exactly what you expected.

  • Worth the Splurge

    Not cheap, and not pretending to be. The cost is the point.

    For instance:A tasting menu with three sommeliers and a chef who's been doing this for thirty years. You'll feel it. You'll be glad.

  • Occasion Worthy

    Save it for the birthday, the apology, the proposal.

    For instance:The river-view dining room that's a bit much on a Tuesday — exactly right on an anniversary.

  • Cheap but Cheerful

    Won't change your life, won't bankrupt you. Honest food.

    For instance:A back-of-the-pub Thai kitchen, paper menus, half the chairs wobble, the green curry is excellent.

  • Paying for the Scene

    You're buying the room as much as the dinner. Know that going in.

    For instance:A rooftop in Shoreditch where the lighting is better than the squid. Worth it if you came for the lighting.

  • Tourist Tax Warning

    Priced for the visitor, not the resident. The locals know.

    For instance:Every pasta place within fifty paces of a major piazza. Including the one with the queue.

  • False Economy

    Cheap, and you'll feel it on the way home.

    For instance:Patty Shack on the El Camino — £14 burger that crashes you out by four in the afternoon.

Section V

Confidence.

Every score in the dispatch ships with a confidence value between nought and one. Below 25 reviews, Harry treats the rating as a rumour. Above 250, he treats it as gospel. In between, he weighs accordingly.

Confidence drops further when critics haven't weighed in, when the opening hours are missing, when nobody's mentioned the place in the last six months. A low-confidence pick can still be a great one — Harry just tells you he's working off thinner evidence so you don't bet the wedding on it.

Section VI

What Harry doesn't know.

  • He hasn't tasted the food. He's working from what other people wrote down.
  • He misses brand-new openings nobody's written about yet — give it a fortnight.
  • He misses the lovely little place down the side street that's never been reviewed. Sorry.
  • He doesn't know about your dietary requirements yet. He's learning.
  • He doesn't read every Reddit thread, every Substack, every group chat — yet.
  • He has no opinion about whether you should have ordered the wine pairing. You should.

Want the receipts? harry@harryhungry.com.

Want to argue? Same address.

— Harry, yours, hungry

← Back to the dispatch