The instinct at a new place is to order what they are known for. I did that for years and it is close to useless as an assessment, because the signature dish is the one they have practised most and the one they will get right on their worst day.
What tells you about a kitchen is the things nobody is watching.
The dal
At any Indian restaurant, this is my first test and it has never misled me.
Dal is cheap, it is on every menu, and nobody orders it to be impressed. Which means a kitchen that makes it well is making it well because that is how they cook, not because they are performing.
What I am looking for: whether it has body without feeling thickened, whether the tempering tastes fresh rather than of old oil, and whether it is seasoned properly rather than just made rich.
A restaurant with excellent dal is almost always good at everything else. A restaurant with thin, flat dal and an impressive main course is usually a place with one good cook and a lot of shortcuts.
Something with a single vegetable
The second test. A simple dry preparation of one vegetable, with nothing to hide behind.
No gravy, no cream, no heap of spices. Just the vegetable, cooked.
You learn immediately whether they can cook. Whether the vegetable has colour on it or has been steamed grey. Whether it is seasoned. Whether it was cooked to order or reheated.
This is also the dish most likely to be genuinely different between two restaurants that look similar on paper, because it depends entirely on technique rather than on a recipe.
The bread, if they make it
Third test, and the most immediate.
Bread arrives fast and is eaten fresh, so there is nowhere to hide. It tells you whether there is somebody competent at the tandoor or the tawa, and it tells you whether the kitchen is under control, because bread is the first thing to suffer when a kitchen is overwhelmed.
Cold bread, or bread that arrives long before or after the food, is a service and timing problem, and timing problems tend to be systemic.
What I stopped paying attention to
Decor, which correlates with price and not with food.
How busy it is, which correlates with location, marketing and the day of the week. I have eaten badly in packed restaurants and extremely well in empty ones.
Online ratings in aggregate, which are dominated by service and value perceptions rather than by cooking. The written reviews are more useful than the score, and only the specific ones.
And the length of the menu, which people cite as a warning sign. In my experience a long menu is a warning sign in some cuisines and completely normal in others, and it depends more on whether the components are shared across dishes.
The signs that do predict
A menu that changes, even slightly, even seasonally. It means somebody is paying attention.
Staff who can answer a question about how something is made. Not a scripted answer — an actual one.
A kitchen you can see or smell. Places where the smell in the dining room is of cooking rather than of air freshener are usually cooking.
And regulars. Not crowds, regulars — people the staff greet by recognition. That is the single most reliable indicator I know of, and it takes no expertise to spot.
The second visit rule
The thing I have become firmest about. A single visit tells you very little.
Kitchens have bad nights. Somebody is off, something ran out, a large table came in at the wrong moment. Judging a place on one meal is judging it on one night's staffing.
What I do now, for anywhere that seemed promising, is go back within a fortnight and order completely differently. The consistency between the two visits is more informative than either visit alone.
Places that are good twice are good. Places that are excellent once and mediocre once are a gamble, and there are more of those than people admit.
The thing this is not for
All of the above is about assessing a kitchen. It is not about whether you will enjoy the evening, which depends on company, timing, mood and whether the room is too loud.
I have had wonderful evenings with mediocre food and miserable ones with excellent food, and pretending the cooking is the whole thing is a mistake I made for a while.
Ordering off-peak
A last practical note. The same kitchen produces different food at different hours.
Right at opening, everything is fresh and the staff are unhurried, and the cooking is often at its most careful. Mid-service at peak, things are faster and rougher. Late in the evening, you are frequently eating what is left.
If I am assessing a place seriously I go early rather than at the busiest hour, on the basis that I want to see what they can do rather than what they can do under pressure.
If I am assessing whether to become a regular, I go at peak, because that is when I will usually be eating there.