Methodology

The 100-point scorecard

Every editorial rating is four numbers. Food is 70 of the 100 points because food is why you are going. Service, ambience and portion-to-price value get 10 each. The total is added up by the software. There is no box for a total; nobody can type one in, round one up, or nudge one.

Category	Points	The question
Food	0–70	Did it blow your mind, or was it merely fine?
Service	0–10	Did the service deliver what this kind of place promises?
Ambience	0–10	Did the room, the noise, the energy help or hurt?
Portion-to-price value	0–10	Did what you got justify what you paid?
Total	0–100	Computed. Never typed.

Scores measure whether a restaurant keeps its own promise, not how much marble is in the room. A taco stand can earn 10 for service if the service is warm, fast and right for a taco stand. A tasting menu gets no free points for being expensive. A taco stand and a tasting menu can land on the same number.

Food anchors, so the numbers mean the same thing from visit to visit: 68–70 is among the most compelling food the editor has eaten anywhere. 59–63 is excellent, a clear reason to return. 40–49 is competent and forgettable. Below 30 the editor would tell you not to go.

The 80 line

We recommend a restaurant only when an editor has visited it, it scored above 80, and the Founding Editor approved it. All three. No exceptions.

Score 80 or below and the restaurant is marked Not Recommended by Our Editors. It stays in the directory with its score and every story we have written about it. It will not appear in recommendation search, best-of lists, The Pass or the newsletter. It can be re-evaluated after a material change or an approved revisit.

Our policy on a first low score: when a restaurant’s first editorial visit lands at 80 or below, the Not Recommended label is held for seven days before it appears publicly. The score is recorded and audited immediately; the hold exists so an editor can check for a data error or a material change before a single visit becomes a public verdict. The hold never applies to a restaurant that already carries a published score.

One visit can qualify a restaurant scoring 81–89. A score of 90 or higher is shown as provisional until a second eligible editor visit confirms it. Flagship placements need that second visit too.

Launch mode: while the Founding Editor is the only editor who records editorial visits, a 90+ score may be confirmed by a second self-paid visit by the same editor at least 14 days later and at a different meal period. Which rule applied is written to the audit log with the score, and the strict two-editor rule returns as soon as a second scoring editor exists.

Directory inclusion is not endorsement. Thousands of places are listed so you can find them by name and see what we know. Search a restaurant by its exact name and you will get its record even if we have not rated it, labeled as such.

Anonymous, self-paid visits

Editorial visits are anonymous. We do not call ahead, we do not announce ourselves, and we pay the check. At launch the Founding Editor pays for his own qualification visits out of pocket. If an editor is recognized or comped, that is recorded on the visit and the rating is flagged; it does not quietly disappear into the average.

Every rating stores who wrote it, their role, the date of the visit, the meal period, and any disclosure. Editors carry more weight in the aggregate than contributors, and that weighting is fixed by role, not by mood.

Receipt verification

Phase 2. When community ratings open, only visits backed by a receipt will count toward a public Community Score, and that score will always sit apart from the Editorial Score, never blended into one number. A public Community Score needs ten verified ratings from ten different accounts. A receipt proves a paid visit happened. It does not prove the opinion is right, and we will say so on every profile.

Until then, restaurant profiles say Not enough verified diner ratings yet. We would rather show an empty box than a number we cannot stand behind.

The commercial firewall

Nothing on this site that looks like a judgment is for sale. Not a score, not a point on a score, not recommendation status, not a spot on The Pass, not a quote. Sponsored content is labeled Sponsored on the card, in the headline area, and in a banner above the text. It is never bylined as reporting and never carries a score.

We disclose gifted access, press previews, events, collaborations, employment, consulting, friendships and ownership interests, on the story and on the editor profile. Where a restaurant has a relationship with anyone here, the disclosure sits on its profile, above the quotes. We do not turn press releases into bylined news. We do not use AI-written prose to simulate a visit, a tasting, a quote or an opinion.

The four dates

A restaurant record never shows a single Last updated, because that phrase hides more than it tells. Instead every profile shows four separate dates:

• Last editor visit — when one of us last ate there.
• Menu verified — when we last confirmed the menu items we track.
• Last community rating — when the most recent receipt-verified diner rating landed (phase 2).
• Record updated — when a factual field (hours, phone, address, booking) last changed.

A dish named in a story shows the date the evidence was gathered and whether it is currently available. A story from last year does not overwrite what the menu says today. Dish availability is fresh for 14 days, a full menu for 30, hours for 7; after that we warn.

Corrections and challenges

When we get a material fact wrong we fix it and log the correction, dated, at the bottom of the story. Every restaurant profile has a Report a change form. It is for facts: closed, moved, dish gone, hours wrong. It is not a rating, and it cannot move a score.

The Pass

A deliberately short list of where to eat in Los Angeles now, dated and signed by the editor who made it. Additions and removals are visible in every edition and on each restaurant profile. Being taken off the list is a dated decision with a note, not a deletion.