By Matt Wampler, CEO of ClearCOGS
The most common reason a restaurant group delays a forecasting project is a problem that forecasting was never going to solve.
I talked recently with a director of operations at a breakfast group in the twenty-unit range, part corporate and part franchised. He is not a man in trouble. Food cost sits in the low twenties and holds there across the system. His worst outlier runs about a point high. The recipes are documented, the yields are defined, the culinary team is disciplined about sourcing and pricing. By almost any measure, this group runs well.
Ask him whether he is ready to hand his kitchen managers a daily prep number instead of a whiteboard, and the answer is no. Not yet. There is work to do on the back end first.
The work is real. It is also, as far as prep is concerned, beside the point.
The Error That Poisons Everything Downstream
Here is the specific thing eating his quarter.
Broadline distributors sometimes ship a split case and a full case under the same item number. Same code on the invoice, different quantity in the box. Somewhere in the back office, a conversion gets set once and then inherited by everything that touches it. The cost per unit is now wrong.
That single wrong number does not stay put. It flows into theoretical usage, because theoretical usage is built from unit costs. Theoretical usage flows into the variance between what the store should have used and what it actually used. That variance is what the group uses to judge whether a location is running tight, and in many groups it is also what determines whether a general manager earns a bonus.
So the unit leaders stop believing the report. They pull the numbers out, rebuild them in a spreadsheet, and only then does anyone feel comfortable paying out. The system produces a figure. The spreadsheet produces the figure people actually use.
You can measure how much a group trusts its own back office by counting the spreadsheets that sit between the report and the decision.
This Is a Known Failure Class, and It Is Not a Restaurant Problem
In 1999, NASA lost the Mars Climate Orbiter after nine and a half months of flight. The Mishap Investigation Board found a single root cause: a ground software file reported thruster impulse in pound-force seconds when the interface specification required newton-seconds. One pound of force is about 4.45 newtons, so every trajectory calculation quietly underestimated the effect of hundreds of small maneuvers. By the time the spacecraft reached Mars, it was roughly 170 kilometers below the planned approach. It did not survive.
The board made a point worth sitting with. Mistakes happen on spacecraft projects all the time, and the processes in place normally catch them before they matter. This one was not caught.
There is a second detail in that report that restaurant operators should find uncomfortably familiar. In the months before the loss, engineers were running more than one method of estimating the spacecraft’s position, and the methods disagreed. One of them consistently indicated a closer approach to the planet. The disagreement was noticed. It was never resolved.
That is the spreadsheet. When a general manager rebuilds the variance by hand and lands somewhere different than the system, the group has just generated a second independent estimate that disagrees with the first. Most groups file that under inconvenience. It is closer to a finding.
So yes, fix the units. It is worth doing and the cost of not doing it compounds silently, which is the worst way for a cost to behave.
Prep and Costing Do Not Read the Same Data
Here is where the sequencing goes wrong.
The group in question treats the back office cleanup as a prerequisite. Get the units right, then the inventory numbers are trustworthy, then the recipes can be trusted, then we can automate prep. It is a sensible-sounding chain. It is also not how the underlying math works, because prep and costing draw on two different sets of data.
Costing runs on master data. Invoices, pack sizes, unit conversions, inventory counts, valuation. Every one of those steps is a place where a bad conversion can enter and corrupt the result. That is the pipeline he is trying to clean.
Prep runs on transaction history. What sold, item by item, at what time, on what day, under what conditions, paired with a recipe that says how much of each ingredient goes into each item. The forecast is counting units, not dollars.
Which means a wrong case size makes your food cost report wrong. It does not make tomorrow’s chicken number wrong. The forecast never asked what the chicken cost. It asked how many portions of it walked out the door last Tuesday, and the Tuesday before that, and the eleven Tuesdays before that, and what was different about the ones that broke pattern.
A group can have genuinely unreliable cost data and completely reliable demand data at the same time. Most groups in this position do, because point of sale data tends to be clean for the same reason it tends to be boring: nobody hand-enters it.
Why the Sequencing Matters More Than the Cleanup
Consider what each project actually costs.
Cleaning up master data across twenty locations and several hundred recipes is a multi-quarter effort with no operational payoff until it is largely finished. A half-cleaned item file produces the same distrust as an uncleaned one.
Putting a real prep number in front of a kitchen manager is a short project that starts paying in week one and keeps paying every morning after. It also happens to build the thing the cleanup project needs most, which is a team that has recent experience trusting a number that came out of a system.
Groups routinely sequence the long project first, stall halfway through it, and arrive eighteen months later with cleaner data and the same whiteboard.
The Part That Actually Threatens Growth
For a franchisor, this stops being a tidiness question and becomes a transferability question.
A kitchen manager with a few years in one building develops a feel for the store. Par levels, batch counts, what a rainy Saturday does to the line. That instinct is real and it is valuable, and it is also completely local. It does not transfer, it cannot be trained quickly, and it takes a new operator months to build while the P&L absorbs the education.
In a home market with regional support nearby, that cost is survivable. In a market three states away where the franchisee has no bench and no history, the group is asking a brand new team to develop institutional knowledge from scratch and hoping the ramp is short. The most common feedback a franchisor gets after a successful out-of-market opening is not that something was broken. It is a question: how do we make this easier?
Handing that team a number is how. And the number does not require the cleanup to be finished.
The Question Worth Asking Instead
When an operator says they are not ready yet, what they usually mean is that one of their data problems has become a blanket veto on everything.
The better question is narrower. Which decisions actually depend on the data that is broken? Costing and variance reporting do. Prep, ordering quantities, and the daily production plan largely do not. Sorting the decisions by what they truly require is usually enough to find something worth starting on Monday.
This is the part of the problem we spend our days on at ClearCOGS: turning the operating data a restaurant already has into the prep, ordering, and labor decisions that get made before the shift starts, without asking the operator to finish a data project first.
If you are in the middle of a cleanup and wondering what you can move on now, Let’s Talk.
Sources
- NASA, Mars Climate Orbiter Mishap Investigation Board Phase I Report, November 10, 1999.
