Inventory systems for small restaurants that are worth the effort

Most small restaurants have tried inventory twice. Once with a spreadsheet somebody built with real care, once with software that came free with the POS. Both died in a walkin at eleven at night with two people disagreeing about what a case is.

The problem was never discipline. The count was designed to produce a valuation, and a valuation is useless to a person deciding how much chicken to order on Tuesday.

Why does full inventory fail in a small restaurant?

Because the work scales with the number of items and the value scales with dollars, and those two lists are nothing alike.

A hundred and forty line count takes two people ninety minutes and produces one number at month end. Twenty items carry most of the money. The rest are on the sheet because they were on the sheet the first time somebody made one.

Price that habit. Two people for ninety minutes is three labor hours a week, or 156 hours a year. At a loaded hourly cost of $22, meaning wages plus payroll taxes plus whatever benefits you carry, that is $3,432 a year spent producing a valuation. Nobody orders differently because of it. The short count costs a fifth of that and changes an order every week.

The second failure is quieter. A count taken at an inconsistent time is not a count, it is a rumor. Sunday after a delivery and Tuesday before one describe two different restaurants, and comparing them produces a variance that exists entirely inside your calendar.

The third reason is that nothing downstream depended on it. The order got placed from a walkthrough and a feeling, and the count got taken afterward for the file. A number with no decision attached stops getting taken correctly within about three weeks, then stops getting taken.

Which items actually belong on the count?

Sort your last month of purchases by total dollars and draw a line after twenty items. That is your list. Add anything stolen, over portioned, or thrown away often, even if it is cheap, because those are the items where a count tells you something an invoice cannot.

Then add nothing else. The temptation to include the rest is where every failed system started.

Three rules for the list you keep:

• Count in the unit you buy and the unit you use. If the sheet says case and the cook says pan, the number is wrong every week.
• Count in walking order, matching the path through the walkin, freezer, and dry storage. An alphabetical sheet doubles the time.
• One person owns it. Two people rotating produce two methods.

Everything outside the twenty gets managed by par and by eye, which is what you were doing anyway, except now on purpose and only where it does not matter.

What makes a count accurate enough to trust?

Same day, same time, same person, same units. That is the entire answer, and it matters more than the software.

Accuracy here does not mean precise. It means consistent. A count that runs slightly optimistic every week still produces a usable trend, because you are comparing your restaurant to itself. A count that is perfect one week and rushed the next produces noise that looks like a food cost problem and sends you hunting for a thief who does not exist.

Same discipline as setting a baseline before you change anything. Measure the same way twice before you believe a difference means something.

The other requirement is that the count feeds the order. Counted, subtracted from par, ordered. If your order is written before the count is finished, the count is decoration.

What does the count tell you that the invoice does not?

Usage. The invoice tells you what you bought. Only the count tells you what left the building.

Purchases plus opening count minus closing count gives you what was used. Compare that against what you sold, and the gap is the whole conversation. Something went out the door that nobody paid for.

That gap has four common causes, and they are worth separating before you act:

• Waste and spoilage, which is measurable without a full program at all. Tracking waste with nothing more than a clipboard will explain more variance in most kitchens than the inventory sheet does.
• Over portioning, which is a training issue and shows up on high volume items first.
• Comps and remakes, where the food left the building legitimately but the sale did not happen. What comps and voids tell you about the floor catches what inventory blames on theft.
• Receiving errors, where you paid for twelve and got eleven.

Theft is not on the list. It exists, but it is the last thing to check. In the restaurants I ran, the number that looked like theft was a portion scoop or a receiving door more often than it was a person.

The same logic that finds the shifts causing four points of labor variance applies here. Do not look for a general condition. Look for the specific item, station, and week.

What does the gap on one item actually cost?

Work one item all the way through and the whole method becomes obvious. Use your own numbers in place of these.

Take chicken thigh, the highest dollar line on most casual menus. One week:

Now the other side. The POS says you sold 1,240 portions that week. The spec is six ounces. That is 7,440 ounces, or 465 lb.

You used 536 and you sold 465. The gap is 71 lb in one week.

At a landed cost of $3.20 a pound, that gap is $227. Held for a year, it is roughly $11,800.

The National Restaurant Association puts the pre-tax margin of a typical independent restaurant near 5%. At that margin, $11,800 of profit is what a restaurant doing roughly $236,000 in annual sales produces. One item, one gap, and the equivalent of a quarter million dollars of sales you would otherwise have to go earn.

Nothing in that arithmetic required software. It required a count on Sunday, a count on the following Sunday, the invoices in between, and one number off the POS.

Now the part people skip. The gap is not the answer, it is the question. Seventy-one pounds is roughly 190 portions of product that left without a sale. Before you conclude anything, check the four causes in order: waste and spoilage, over portioning, comps and remakes, and receiving errors. In the restaurants I ran, a gap that size was almost always a scoop, a scale, or a receiving door.

Run one item for four weeks before you act. A single week's gap can be a counting error at either end. Four weeks pointing the same direction is a condition.

What is your food cost supposed to be?

Nobody credible will tell you, and the confidence of anyone who does is the tell.

The rule of thumb figures that get handed to operators, the ones that say your food cost should sit somewhere between 28% and 32%, do not travel with a sample size, a segment definition, or a date. They are folklore with a decimal point.

The best available data on restaurant cost structure is the National Restaurant Association's Restaurant Operations Data Abstract, 2025 edition, built on financial and operating data from more than 900 restaurant operators nationwide. It does break out food cost by average check, annual sales volume, size of community and location. Those cuts sit behind a paywall, which is why no segment food cost standard appears in this article. And the Association attaches its own caveat to everything in the publication: the data is "not intended to represent standards or goals for individual restaurants," and is meant as a management tool for gauging performance.

This pattern repeats across industries and it is worth recognizing. In dentistry, the figure "$850 per new patient" circulates on hundreds of vendor pages, and the American Dental Association's Health Policy Institute, which is the correct primary source for dental practice economics, publishes no such number. Every circulating value traces back to another vendor page. A benchmark that cannot name its sample is not a benchmark, it is a sales aid.

What the Association does publish is direction, and the direction matters. Average wholesale food prices are up 35% since 2019 and total expenses for an average restaurant rose 36% between 2019 and 2026. A target set from a rule of thumb five years ago describes a restaurant that no longer exists.

So compare your count to your own last eight weeks. That is the only ruler in this article that is actually calibrated to your building.

Where does the short count break down?

Three places worth naming before you build the sheet.

When the recipe is fiction. Theoretical usage is calculated from your spec. If the spec says six ounces and the line has been sending seven for a year, your variance is real and your explanation for it will be wrong. Weigh five plates off the pass before you trust a single number in the table above.

When the gap is upstream of the walkin. You can count perfectly and still be short because you paid for twelve cases and received eleven. Receiving is a different control with a different fix, and no amount of counting reaches it. The count tells you something is missing. Only a signature at the back door tells you it never arrived.

When the top twenty churns. A seasonal menu reshuffles the dollar ranking every quarter. Rerank the list every twelve weeks rather than defending last spring's sheet, and expect two or three items to swap out each time.

What to do this week

Rank the last four weeks of invoices by total spend and take the top twenty. Half an hour, and it is the only setup step that matters.

Build a one page sheet in walking order with two columns, case and each. Name one counter and pick one day. Sunday night or Monday morning works for most kitchens because delivery patterns make it repeatable.

Count for three weeks before you draw a conclusion. The first count is a starting point, the second is a comparison, the third is the first one you should believe.

On week four, calculate usage on your five biggest items and compare it against what you sold. Take the largest gap and work that one item until you know why.

Do not buy software until you have done this by hand for a month. You will know exactly what you need it to do, and you will find that a shared spreadsheet does most of it.

Questions about what to count?

Email me at eric@seod.com with the number of items on your current inventory sheet and how long a count takes you. That is all I need. I will tell you which items I would cut, what the short sheet looks like for your kind of menu, and which day I would count on.

Sixteen years running multi unit restaurants and more than $54M in annual P&L means I have built these sheets, watched them get abandoned, and rebuilt them shorter. The short one survives.

Otherwise, there is more on running the operation here.