Why my food cost was so high, and how long it took me to find out
When I stopped believing my own number
There was a stretch where the group food cost number moved period to period by more than anything on the floor could account for. Same menu, same vendors, same volume, same people. A swing like that in four weeks is not an operating problem. It is a measurement problem wearing an operating problem's clothes.
So I did the thing I now tell everyone to do first, except I did it far later than I should have. I stopped trying to fix the number and went and looked at where the number came from. I pulled a quarter of invoices. I stood in walk-ins with a clipboard on a Sunday. I asked every manager who owned a count to describe how they count, separately, and I wrote down what each of them said.
The answers did not match each other. That was the whole finding. My food cost was not high. My food cost was unknown, and unknown always rounds up.
Four gaps where my money was actually going
None of these are exotic. That is the point. Every one of them is boring, and every one of them was open for months while I was looking at menu engineering.
Gap 1
Between the invoice and the walk-in
Product got signed for at the back door by whoever was closest to it. Not weighed, not spot-checked against the packing slip, sometimes not even the right item. Put a scale at the door and you will find cases that do not weigh what the invoice says they weigh. That is not theft. That is nobody being responsible for the receiving door at 6:15 in the morning.
Gap 2
Between the prep sheet and the pan
My prep sheets said one number. My cooks made what they felt like the day needed, which was always more, because running out is humiliating and over-producing is invisible. Nobody has ever been yelled at for making too much. That asymmetry alone was costing me points.
Gap 3
Between the count and the count
Two sites counted the same walk-in on different days, using different sheets, in different units. One counted cases, one counted pounds. When I rolled them up, the group number was fiction. I was making decisions off fiction for most of a year.
Gap 4
Between the number and the meeting
Every week I sat in a room where a food cost number got read out loud and nobody defended it. Silence reads as agreement. It is not agreement. It is a room full of people hoping the topic moves on.
The argument I stopped having
I want to be specific about portioning fights, because losing a few of them changed how I work. The setup is always the same. A chef is carrying the worst protein variance in the group. I have the data. I bring it to them. They listen for about two minutes and then walk me into their kitchen to show me the thing the data cannot see.
More often than not, what they show me is that the spec no longer matches the floor. Pan sizes change. Suppliers change cut and trim. Somebody swapped a piece of equipment for an unrelated reason and nobody went back and rewrote anything. The cooks are filling the pan, which is what cooks do. The spec is asking for a yield the pan cannot produce.
I have spent whole meetings arguing with people about their discipline when the actual defect was a document I owned. Those are the right arguments to lose.
Since then, the first question I ask about any variance is not who. It is what is the standard, when was it last written, and does the equipment on the floor still match it. Most of the time the standard is stale and the person is fine.
Five things I got wrong
I am not going to pretend I saw this cleanly. I did most of the wrong things first, in roughly this order.
- I chased the number instead of the inputs. I would see a bad percentage and start cutting the same week, before I knew whether the percentage was even real.
- I treated it as a chef problem. I have argued portioning with chefs who were dead on spec, when the defect was a spec nobody had rewritten since the equipment on the floor changed.
- I bought software before agreeing on the count. New app, same people counting different ways. Faster wrong numbers.
- I ran it as a push instead of a cadence. Push gets you a good stretch, then you get pulled onto a labor problem and the whole thing decays inside a month.
- I never made one person own it. Everyone was involved. Nobody was accountable.
The last one is the expensive one. For a long stretch, food cost was a shared concern, which is another way of saying it was nobody's job on any given Tuesday. Shared concern is how a number drifts for three quarters while everyone in the room can honestly say they were worried about it.
What I would do differently, in order
If someone dropped me into a group tomorrow with a food cost they did not trust, I would not touch the menu, the vendors, or a single portion for six weeks. Here is the sequence I would actually run, and I would run it in this order even when it felt slow.
Week one, count it myself. Not review a count. Physically walk every walk-in and dry store with the manager who owns it and count against what the system says should be there. I am not looking for the variance. I am looking at how they count, what they skip, what they estimate, and what unit they think in. The gap between sites is the finding.
Week one, also pull invoices. Twelve weeks of them, reconciled against the food cost line and the inventory change. Not to catch a vendor. To find out whether the paperwork can even support the number. Usually it cannot, and usually the reason is a credit memo process nobody follows.
Week two, find the clean site. In every group I have run or looked at, one location holds its number. It is never because they have better people. It is almost always one habit that manager kept doing after everyone else stopped. Find that habit and you have your standard, already proven inside your own operation, which is worth ten times a habit imported from outside.
Weeks two through eight, one cadence, unchanged. Same count day, same sheet, same units, same meeting, same person running it, same agenda. And a rule I would enforce hard: the number gets defended out loud by the person who owns it, or it gets fixed. No silence. Silence was the single most expensive thing in my variance meetings for years.
Only then, decide anything. After six weeks of a cadence you trust, you will find that half the things you were going to change do not need changing, and the two that do are obvious to everybody in the room. That is a much easier conversation than the one where you cut portions and hope.
What it actually took
Where this ended up: food cost holds between 28 and 35 percent depending on the model. Institutional and contract catering carries a different line than the retail cafe, central production a different one again, and each of them holds its own. That is the part worth saying clearly, because one number across four different businesses would not mean anything. Food waste came down 14 percent over nine months and margin improved about 20 percent. Not in a week. The rebuild of receiving, counting, production planning and the weekly review had to land first, and rolling MarketMan across all 14 sites is what finally made one count mean one thing everywhere. Trusting the number was the real deliverable. Everything else was a consequence of it.
I would also say plainly that the first three weeks looked like nothing was happening. Costs did not move. There was no visible win to report. If you are going to do this properly, tell your ownership up front that the first month buys you a real number and not a better one, because if you do not say that in advance you will get pulled off it in week three the way I have been pulled off it before.
If you only take one thing
High food cost is almost never a kitchen behaving badly. In my experience it is a control system that stopped being maintained, and a set of standards that quietly went out of date while everyone kept working hard against them.
Before you cut anything, go find out whether your number is real. Most of the time that is the whole job, and the rest follows from it.
Related reading
If you want this fixed in your operation
Everything above is what I learned running it myself, and you are welcome to run it yourself. Doing this inside someone else's operation is the work I do through XenoSoft Solutions, and that side of it is written up there rather than here.