Restaurant analytics: what the numbers should tell an owner, and how often
Most owners already have more numbers than they can read. The POS prints a sales summary. The scheduling app has a labor screen. The bookkeeper sends a P&L three weeks after the month closed. That is a lot of data and not much analytics.
Analytics is the part that comes after the numbers. Which ones matter, what normal looks like for your restaurant, and what you are going to do about the one that moved. This page is about that part. If you want the tools and their prices, they are in our guide to restaurant reporting software.
Start with one question
Every number worth your time answers one question: did we have a good night, and if not, why?
Sales alone cannot answer it. A big Saturday with three extra servers on the floor, a comp-heavy table and a case price that went up last week can look great on the sales line and still leave less in your pocket than it should have. So the first number is not sales. It is what you kept, after every bill: food, labor, rent and the fixed bills.
Rent is monthly, so how do you get a nightly number? Take the month’s rent, insurance and the other fixed bills, and divide by the nights you are open. That is each night’s share. Everything else on the page explains what is left.
What restaurant analytics should show you every morning
Six numbers. These move night to night, and you can still act on them today.
- What you kept. Sales minus every bill: food, labor, and that night’s share of rent and the fixed bills. The other five explain this one.
- Sales against your own normal. Not against last night. Against the same weekday over recent weeks. A Monday is not a Saturday, and comparing them teaches you nothing.
- Checks and average check. Sales went down. Was it fewer people or smaller checks? Fewer people is a traffic question: weather, an event across town, a slow week. Smaller checks is a floor question: drinks, appetizers, desserts, who is suggesting what.
- Labor by job, against the sales it produced. Not one labor percentage for the whole night. Kitchen, servers, bar, hosts, each one. Who is getting close to overtime this week, while there is still time to fix the schedule.
- Voids and discounts. By the hour and by server, against the house normal. Most of the time it is nothing. A server training, a slow kitchen, a birthday table. Sometimes it is not nothing, and you want to know on Tuesday, not in March.
- Kitchen ticket times. If your kitchen runs on a screen, look at times by the hour and the slowest station. A kitchen that falls behind at 7:30 is a table that waited, and guests remember.
If it takes longer than your coffee to read them, the report is too long.
Every week: the three that need a few days
Some numbers are too noisy to read nightly. A week smooths them out.
- Prime cost. Food and beverage cost plus labor, as a share of sales. Use invoices for the food side, the ones that arrived that week. Count the walk-in once a month to correct it. How to work it out by hand.
- Invoice prices. Put the same item from the same vendor side by side, week over week. Prices rarely jump. They creep, a little at a time, and nobody calls to tell you. Any one invoice, you would sign. Four in a row, side by side, is a call to your rep.
- Menu mix. What sold, what did not, and what each plate costs you now, not what it cost when you priced the menu. An item that sells well and costs more than it used to is the one to look at first.
Every month: the P&L, closed fast
The monthly P&L is still the scorecard. The problem is timing. A P&L that lands on the twentieth of the next month is history. By then most of another month of the same problem is already in.
Ask for the month closed in days, not weeks. Read every line against your own normal for that line. Then, and only then, against the industry.
For that check, one published split is worth knowing. The National Restaurant Association’s Restaurant Operations Data Abstract, 2025 edition, reporting 2024 figures from more than nine hundred restaurants, found that full-service operators who made a pre-tax profit ran wages at 34.2% of sales, and operators who lost money ran 42.9%. Count labor the way the Association does, benefits included, not just the hourly rate. If yours sits nearer 43 than 34, that is where to look first. The same report puts the middle full-service restaurant’s pre-tax profit at 2.8% of sales.
One caution. The restaurant in the middle is not your restaurant, and figures from separate lines do not add up into one. More on that here. Use the industry as a check on the month, not a target.
Restaurant analytics starts with your own normal
This is the idea that matters most, whatever you use.
An average tells you what a restaurant somewhere did. Your own normal tells you what your restaurant does on a Tuesday in November, with your menu, your street and your team. When labor moves three points against your own recent Tuesdays, something changed, and you can go find it. When you are a point off an industry figure, you have learned very little.
Good analytics builds that normal from your own history, and it gets sharper the longer it watches. After a couple of months it should be able to tell a normal quiet night from one that went wrong. Ask any tool how it works out normal. If the answer is "the industry average," keep looking.
What to do when a number moves
A number that moved is a question, not an answer. Three steps, every time.
1. Make sure it’s real. Does the sales total match the POS close? Is a source missing? Good analytics says so on the page, in the place the number would have been, instead of filling the gap. 2. Find the why. Weather, an event, a menu change, a new server, a vendor price. The manager’s notes from the shift answer half of these. A sentence at the end of the night saves an hour of guessing in the morning. 3. Decide one thing. Move a shift. Call the rep. Talk to the server. Re-cost the burger. If nobody does anything differently, the analytics did not work, however good the charts looked.
What AI adds to restaurant reporting, and what it doesn’t
AI is the newest word on every restaurant software page, so here is the plain version.
What it is good at: you ask in your own words, "How did lunch do last week against the same week last year?", and you get the answer with a chart. No digging through report menus, no exporting to a spreadsheet. It can also write the story under the numbers: what moved, why, and what to look at first.
The catch to watch for: some AI tools guess when they don’t know. Ask any tool what it does when a number is missing. The right answer is that it says so, not that it fills the gap with something that sounds right. And ask to see the rows behind an answer. If it can’t show you, don’t bet a schedule on it.
What it can’t do: it doesn’t know your regulars, your staff, or that the walk-in was down on Tuesday. It points. You decide.
How Syphor handles it, since it is ours: it answers only from your own numbers, with the rows it used underneath. When it can’t see something, it says so. And it is read-only. It cannot change a price, void a check or move a shift.
Kinds of restaurant analytics software
Short version. The full list, 13 tools with every published price read on each company’s own website on November 11, 2026, is in our guide to restaurant reporting software.
- Your POS’s own reports. Start here. Your POS already has every check and every hour, and most owners use only part of what it can show.
- Dashboards. A screen you open and study. Avero’s Starter plan was $0 when we read its site on November 11, 2026, and it is worth a try if most of your questions are about sales.
- A daily summary that comes to you. OpSage, Livelytics, Syphor and Dayfolio all send one. Syphor is ours: one page before seven, by email or text, that reads your POS, your invoices, the weather, your reviews and your managers’ notes.
- Invoice and back-office systems. MarginEdge is built on invoices and food cost. Restaurant365 keeps the books, inventory, labor and payroll together. The right first stop when food cost, not sales, is what keeps you up at night.
Published prices, read November 11, 2026, started at $0 and went up to $350 a location a month, and some add a setup fee.
Three questions owners ask
Is restaurant analytics worth it for one restaurant? It can be. The six morning numbers above matter just as much at one restaurant as at twenty. Start with your POS reports. Spend more when you catch yourself putting invoices, the schedule and the weather side by side by hand.
How is analytics different from reporting? A report tells you what happened. Analytics tells you whether it was normal, why it moved, and what to do. Same numbers, one more step.
What should I ask before I sign? Start with two questions. How does it work out my normal? Does it say so when data is missing? Then ask who does the setup and when the first report arrives. If setup turns into your homework, that tells you how the next year will go.
Syphor reads your point of sale, your invoices, your hours, your kitchen and your manager’s note, sets every line against that restaurant’s own normal, and sends one page before seven. $199 a restaurant a month. Thirty days on your own numbers first. Book a demo.
Or skip the demo — tell us what you run and we set it up, or call 1 888 593 3881.
SyphorOne page, every morning.
Read a whole morning — one night, every lens. Or five pages, as a PDF, for somebody who has to see it and was not on the call.