Reading your POS data for something other than sales

The daily sales number is the least useful thing in the system. It tells you what happened and nothing about why. Everything that would let you change next Tuesday sits two menus deeper, in reports that shipped with the software and have never been opened.

You already paid for this. It is sitting in the machine.

What does item mix tell you that the sales total does not?

It tells you which items are carrying the restaurant and which ones are occupying the menu.

Sort every item by units sold and by total dollars, then read the two lists side by side. Four groups fall out. High units and high margin is your business. High units and low margin is the trap, because it feels busy and pays badly. Low units and high margin is worth promoting. Low units and low margin is why your prep list is long.

That fourth group is the one to act on. Every slow item still costs a spot in the walkin, a line on the prep list, and a chance for a cook to make it wrong because they make it twice a week.

A menu item that sells rarely is not free just because nobody orders it.

Item mix also drives pars. If you set prep pars from memory, the report will show you at least one item prepped for a demand pattern that moved. And when you go to update the menu everywhere it exists, remember the public copy counts too, because how the menu page on your website is structured is what people and search engines actually read.

Why does sales by hour change how you schedule?

Because a daily total hides the shape of the day, and the shape is what you staff.

Two Wednesdays with identical sales can be completely different problems. One arrived in a two hour block, the other spread flat across seven hours. The first needs bodies stacked into a window. The second needs fewer people, longer. Same total, opposite schedule.

Pull sales by half hour for four weeks and average by day. What usually shows up:

• A rush that starts earlier than the schedule assumes, so the first thirty minutes are covered by people still setting up
• A dead hour in the middle that nobody has ever cut
• A tail at the end where four people are closing around two tables

Then compare that curve to your posted schedule. The gap between the two is the money, and it is the same gap that produces most of the scheduling mistakes that quietly cost a shift.

Do this by day of week, never in aggregate. Friday and Tuesday do not share a shape and averaging them creates a curve that describes neither.

What does one report actually change?

One report, one afternoon, and a number you can act on. Here is the arithmetic with figures you can swap for your own.

Pull sales by half hour for a Wednesday. Total for the day, $3,900. Total scheduled labor, 52 hours.

Now look at the window from 2:30 to 4:30. Sales in those two hours: $310. People on the clock: five. That is 10 labor hours producing $310, or $31 of sales per labor hour.

The rest of the day did $3,590 across 42 hours, which is roughly $85 of sales per labor hour.

The daily total told you none of that. The daily total said $3,900, which is a fine Wednesday.

Cut two people out of that window and you save four labor hours a day. At a loaded hourly cost of $21, meaning wages plus payroll taxes plus whatever benefits you carry, that is $84 a day. Six days a week is $504. A year is $26,208.

The National Restaurant Association puts the pre-tax margin of a typical independent restaurant near 5%. At that margin, $26,208 of profit is what a restaurant doing roughly $524,000 in annual sales produces. One report, read once, worth more than half a million dollars of sales you do not have to go find.

For context on the size of the prize, the Association's Restaurant Operations Data Abstract, 2025 edition, built on financial and operating data from more than 900 restaurant operators nationwide, puts salaries and wages including benefits at a median of 36.5% of sales for full service restaurants in 2024, with profitable operators at 34.2% and loss making operators at 42.9%. That is 8.7 points between the two, and windows like the one above are how the gap is assembled.

Before you cut anything, check the window twice. Some dead hours are prep hours wearing the wrong hat, and cutting them moves the cost into a slower dinner rather than removing it.

Why is an average the wrong summary?

Because an average is designed to hide the thing you are looking for, and this is the most common way a report gets read wrong.

There is a clean illustration of this from speech recognition, of all places. Deepgram, which sells transcription, published an explainer walking through a transcript with a 25% word error rate that was perfectly usable for classifying what a call was about, except that it mistranscribed the single word that mattered, turning "declined" into "designed" and silently misrouting the call. Their conclusion is the useful part: "when a speech recognition API fails to recognize words important to your analysis, it is not good enough, no matter what the WER is." The industry's fix was to stop reporting the average and start reporting errors on the entities that change meaning.

Your ticket time report has the same disease. An average ticket time of eleven minutes across a Saturday is a number that cannot be wrong and cannot be useful. Inside it are three hundred tickets at nine minutes and forty tickets at twenty-six minutes, and the forty are the ones that produced the comps, the bad reviews, and the table that will not come back.

So read distributions, not averages, wherever the report allows it. Concretely:

• Ticket time: the slowest tenth of tickets, and what hour they landed in
• Sales by hour: the worst two windows, not the daily total
• Item mix: the bottom fifteen items, not the top five
• Discounts: the largest single entries, not the monthly sum

Every one of those is the same move. Sort to the tail, because the tail is where the decision is.

What can the POS not see?

Demand it never recorded, and this is the honest limit of the whole exercise.

Your POS is a record of transactions. A guest who called at 7:10 on a Friday, got a busy line, and ate somewhere else appears nowhere in it. Neither does the party of twelve that did not book because nobody picked up.

There is real data on how large that hole is. Maple, a company selling restaurant phone automation, published an analysis of 1.2 million calls across more than 1,000 US restaurant and local business locations between December 2023 and November 2025. 68% of all calls land during the lunch and dinner rushes, 28% at lunch and 40% at dinner. 58% of reservations still originate from phone calls rather than apps or online booking, and reservations are 38% of all calls. This is vendor research, published by a company that sells the fix, and Maple states that some of its figures, including its missed call estimates, are merchant self reported during onboarding rather than platform measured.

Take the measured parts and the point stands. The phone peaks exactly when the floor is least able to answer it, and every one of those calls is demand your sales by hour report will never show you, because it arrived and left without touching the POS.

So when the curve says the 7pm hour is your ceiling, hold that conclusion loosely. It may be your ceiling. It may be the point at which you stopped being reachable.

What do the employee level reports actually show?

Behavior, mostly, and it is the most sensitive data in the building.

Discounts and voids by employee, sales per labor hour by server, items per ticket, ticket time by station. Read all of it as a question rather than a verdict.

Discounts are the clearest signal. One person comping far more than the rest is worth understanding, and the answer is usually not theft. It is often a bad section, a slow station, or somebody never taught what to do when a guest complains. What comps and voids tell you about the floor is the longer version, and the conclusion holds: the number describes a system before it describes a person.

Items per ticket is the honest measure of a server. Not total sales, which mostly reflects the section they were given. Compared across people on the same shift, it tells you who is actually selling.

One rule before you use any of it. Tell the crew you look at these reports and what for. Data that arrives as a surprise accusation stops being useful, because behavior changes to protect against the report rather than to serve the guest.

Which reports are not worth your time?

Anything you cannot act on this week.

Year over year comparisons on a single item. Hourly labor watched in real time, which mostly produces anxious cuts at six thirty. Any dashboard with more than eight numbers on it. Custom reports built for a question you no longer have.

The test is simple. If pulling the report changes something you do in the next seven days, pull it. If it only makes you feel informed, it is entertainment. The same test separates useful marketing measurement from vanity measurement, which is the argument in measuring AI search visibility honestly.

One more trap. Most reports let you choose a date range, and most operators choose the one that flatters the answer they already believe. Pick the window before you pull, not after.

What to do this week

Open your POS back office and find five reports: item mix, sales by hour, discounts by employee, ticket time, and average check with party size. Write down where each one lives. That map is half the work and nobody ever makes it.

Pull item mix for the last ninety days. Sort by units. Look at the bottom fifteen items and ask what each one costs you to keep.

Pull sales by half hour for your two busiest days and your slowest day. Draw the curve next to your schedule.

Pick one thing to change. One item off the menu, or one shift start time moved by thirty minutes. Then check the same report in two weeks.

Put a recurring thirty minute block on your calendar for the same day each week to do this. The habit is worth more than any single report, and it pairs with knowing which numbers to look at before you open each morning.

Questions about what your POS is hiding?

Email me at eric@seod.com with the name of your POS and one screenshot of your item mix report. I will tell you exactly which reports in that system are worth a weekly pull, where they live in the menus, and which two numbers I would put on a page for your managers.

Sixteen years running multi unit restaurants and more than $54M in annual P&L across several different systems, most of them worse than yours. The reports were always there. Nobody had ever been shown where.

Otherwise, there is more on running the operation here.