How to find out what's actually selling — and catch a decline early

Every menu has at least one dish that looks like a star on the surface and one that looks unremarkable and isn't. A sales report already knows which is which — most restaurants just never build the report that shows it.

This is the version that needs nothing beyond what you already have: units sold and revenue earned, from your own billing data. No plate costing, no recipe sheet — that deeper version exists too, and it is worth doing once this one is a habit.

Two lists that disagree, and a third that matters more

Rank every dish two ways over the same period: units sold, and revenue earned. They are not the same list, and the gap between them is where the useful information lives.

• By quantity — how many plates left the kitchen. Tells you what to prep.
• By revenue — how many rupees the dish generated. Tells you what to protect.

A dish can rank near the top of one list and the middle of the other simply because of price. Neither ranking is more "correct" — they answer different questions, and reading only one is the single most common mistake in this kind of analysis.

There is a third signal most restaurants skip entirely: how the dish is trending against a few months ago. A dish can sit comfortably mid-table on both lists this month and still be worth an urgent look, if it sold a third fewer plates than it did last quarter. Quantity and revenue tell you where a dish stands today. Trend tells you where it's heading — and it is the only one of the three that gives you time to act before a dish is simply gone from the menu.

A worked example

One month's sales for eight dishes on a casual dining menu, against the same dishes three months earlier:

Reading quantity alone, butter naan looks like the menu's biggest success and mutton rogan josh looks unremarkable, sitting fifth from the top on volume. Reading revenue, the picture flips: rogan josh is the third-highest earner on the menu despite being fifth by volume — and it is falling fast enough that it will not stay there.

That is the dish this report exists to catch. Down 32% over a quarter while still ranking in the top three by revenue is not a dish to quietly let fade — it is one to ask the kitchen about this week. A recipe substitution, a supplier change in the mutton, a price rise that made it feel expensive next to the biryani, or a cook who left and took the version people liked with them are all more likely explanations than guests simply losing interest, and every one of them is fixable if caught now rather than in another quarter.

Veg manchurian, down 42%, is a smaller problem in rupee terms — it was never a big earner — but the size of the fall is worth the same five minutes of asking why, since a fall that steep rarely happens without a specific cause.

The new dessert has no history to compare against and should not be judged yet. Thirty units in a first month is not a verdict either way.

Reading a decline properly

A falling number is a question, not a conclusion. Before assuming a dish has simply gone out of favour, check what changed on your own side first:

Did the recipe change? A supplier substitution, a portion-size adjustment, or a new cook's version of the same dish can shift how a regular item is received without anyone deciding to change it on purpose.

Did the price move? A dish that quietly got repriced upward, especially relative to a similar item nearby on the menu, can lose volume for a reason that has nothing to do with the dish itself.

Did it move on the menu? A dish relocated to a different page or section, or pushed below the fold on a printed card, sells differently through visibility alone — this is covered from the pricing and placement side in menu engineering.

Is it seasonal? A dish tied to a season or a festival period will always look like it is declining in the months after — compare it against the same period last year, not the month before, to tell a real fall from a predictable one.

Only once those are ruled out is "guests want it less than they used to" the honest answer, and even then it is worth knowing before deciding what to do about it.

What to do with each finding

A simple rule, without needing cost data at all:

• High on both lists, flat or rising — protect it. Never let a dish carrying this much of the menu's identity get quietly shrunk, discounted into thinness, or moved to make room for something new.
• High on quantity, lower on revenue, steady — leave it. This is usually a cheap anchor dish — a side, a starter — that brings people to the table rather than the thing they came for; cutting it on revenue rank alone is a common and avoidable mistake.
• Mid-table but falling sharply — investigate this week, before it falls further. This is the case a single month's snapshot would never catch.
• Low on both, and has been for a while — a genuine candidate to cut, once you've checked it isn't a small group's specific reason for choosing you.
• New, no baseline yet — give it four to six weeks before judging it either way.

Common mistakes

Comparing this month to last month only. A single-period comparison cannot tell a blip from a trend; use a rolling few months.

Judging a new dish too early. The first fortnight undersells almost everything.

Ignoring seasonality. A dish will always look like it is declining right after its season ends — the honest comparison is the same period last year.

Not checking your own changes first. A recipe, price or menu-position change on your side explains most declines that get blamed on fading demand.

Stopping at the report. The list is the start of one conversation with the kitchen, not the end of the exercise.

Taking it further

This version runs on sales data alone, which is exactly why it is the right place to start. The deeper version brings in plate cost, ranks every dish by what it actually contributes in rupees after ingredients, and sorts the whole menu into four groups with a specific action for each — covered fully in menu engineering, which also links to costing a dish properly in how to price menu items. Do this simpler pass first if you have never looked at a best- and worst-seller list at all; it needs nothing you don't already have.

The short version

• Rank dishes two ways — by units sold and by revenue earned. They routinely disagree, and both are correct answers to different questions.
• Add a third signal: trend against a few months ago. It catches a decline while there is still time to act.
• None of this needs plate-cost data — it runs entirely on numbers you already record.
• Never judge a new dish before four to six weeks have passed.
• Check your own recipe, price and menu-placement changes before assuming a falling dish has simply gone out of favour.
• Cut only what stays low on every measure after those checks — and only once you've confirmed it isn't a small group's reason for coming in.