Two Questions: Evaluating a Restaurant and Recommending One

The Conflation

Most restaurant criticism conflates two distinct activities: evaluating a kitchen and advising a traveller. These are related — a poor evaluation should probably not generate a positive recommendation — but they are not the same question, and they do not have the same answer even when the evaluation is clear.

The evaluation asks: is this restaurant doing what it set out to do, and is what it set out to do worth doing? This is a question about the restaurant on its own terms.

The recommendation asks: given who you are, what you want, where you are, how much time you have, and what else is available, should you go here? This is a question about the fit between a specific person and a specific restaurant in a specific context.

These questions have different inputs, different methods, and frequently different answers. Conflating them produces criticism that is unreliable in both directions: it recommends restaurants that are good but wrong for the reader, and declines to recommend restaurants that the reader would benefit from visiting.

The Problem with Evaluation-Only Criticism

Pure evaluation produces rankings. Rankings answer “which restaurant is better than which other restaurant” on some composite scale. This is a coherent exercise, but it is not the same as advice.

A restaurant can be excellent — genuinely excellent — and wrong for a given visitor. The price may be prohibitive. The format (long tasting menu, communal seating, late reservation only) may be inconvenient. The particular quality the restaurant offers may not be the quality the visitor is seeking. None of these facts are captured in an evaluation of the kitchen.

Conversely, a restaurant that receives a middling evaluation may be precisely right for a certain visitor. If you are travelling for two days with a preference for understanding a city’s social life rather than its ambitious cooking, a highly-rated tasting menu restaurant is probably the wrong choice. The neighbourhood boteco that the evaluator rated as merely competent may be the most useful meal you eat.

The Problem with Recommendation-Only Criticism

Pure recommendation — without evaluation — produces the guide as a list of addresses organised by category. This format is useful for logistics but analytically thin. It answers “where can I eat X?” but not “is X at this address worth understanding, and why?”

The deeper problem with recommendation-only criticism is that it mistakes the traveller’s interest for the reader’s interest. A guide that is optimised for the person who visits a city once and needs to maximise a short window is different from a guide that is trying to help a reader understand what a city’s food culture is and why it is the way it is. The first is a logistics document; the second is what this publication is trying to make.

The two serve different needs and should not be confused. A visitor who wants practical logistics needs an itinerary, not an essay. A reader who wants to understand São Paulo needs analysis, not a restaurant list.

Opportunity Cost Is the Mechanism

The mechanism that connects evaluation to recommendation is opportunity cost. A recommendation is not simply “this place is good”; it is “this place is worth the time, money, and alternatives you are giving up to be here.”

This calculation changes radically by context. For a visitor with three days in São Paulo, opportunity cost is very high: every long tasting menu dinner is three neighbourhood meals not eaten. For a resident who eats out regularly, opportunity cost is much lower: the long tasting menu dinner is one occasion among many, and the comparison class is the same city’s restaurants over time.

Most restaurant criticism is written for readers whose context it does not know, and it therefore embeds assumptions about opportunity cost that may be wrong for most readers. The honest critical act is to make the opportunity cost calculation explicit: who is this recommendation for, and what are they giving up?

Confidence Is Not Uniform

A further distinction: evaluation and recommendation carry different confidence levels even from the same evidence base.

An evaluation based on several visits can be stated with reasonable confidence: the kitchen is consistent, the cooking is at this level, the philosophy is this. The evaluation is about the restaurant’s properties.

A recommendation is about fit, and fit depends on variables the critic cannot fully know: the reader’s other options, their time constraints, their prior experience, their specific interests. A recommendation therefore carries lower confidence than an evaluation from the same evidence, and should be stated with appropriate hedging.

This publication’s practice is to separate the evaluation (the six criteria, assessed against the restaurant’s own terms) from the recommendation (a tier, a confidence level, an explicit statement of audience and opportunity cost). The separation is not bureaucratic; it is the only honest response to the fact that these are different questions.

The Instruction

When you read a restaurant recommendation — in this publication or elsewhere — it is worth asking: is this person answering the evaluation question or the recommendation question? If the criticism is conflating them, which direction is the error running? Is an excellent restaurant being recommended to you when it may not be right for your context? Or is a merely competent restaurant being dismissed when it might be exactly what you need?

The instruction is not to distrust recommendations but to understand what they are: an answer to a question about fit, not a verdict about quality. The verdict about quality is a prerequisite, but it is not sufficient. Sufficiency requires knowing who is asking.