By Matt Wampler, CEO of ClearCOGS
Most restaurants can absorb a bad prep day. You over-produce on Tuesday, and the proteins that did not sell fold into Wednesday’s prep cycle with some adjustments. It is not a clean outcome, but it is a recoverable one.
Artisan bakeries do not have that option.
When everything is made fresh and nothing carries over, every production decision is a one-way door. The 20 almond croissants that did not sell on Tuesday are not Tuesday’s prep credit toward Wednesday. They are Tuesday’s loss, settled at closing time. Multiply that across a full case across a full week across multiple locations, and the waste-stockout equation becomes the central operating challenge of the whole business.
What the Target Actually Is
Framing the problem as “waste reduction” misses half of it. The goal for a zero-carry-over operation is not to minimize what goes in the bin. It is to end each day with a case that is near-empty without having been empty at noon.
Those are different targets. And threading the needle between them requires something more precise than a safety margin.
A display case that runs empty by early afternoon signals something specific to the customer: demand outran supply. For a high-quality artisan product where the purchase is often part of an experience, that signal matters. Regulars who came for the specific item they had last Saturday and find it gone before lunch do not simply choose something else. They note it. In some cases they do not come back at the same time, or at all. The revenue miss is visible. The brand erosion is not.
Over-producing to protect against that risk has its own cost. End-of-day waste does not appear as a line item tied to a specific decision. It gets absorbed into food cost variance, and because no single day’s waste is dramatic, the pattern can run for months before anyone attributes it to a structural problem in production planning.
The tightrope is not between good and bad days. It is between two expensive failure modes that are both invisible in the short run.
Why Intuition Stops Transferring at Scale
At one location with a long-tenured baker, the production question gets solved through accumulated pattern recognition. She knows what the Saturday morning rush looks like in April versus October. She knows what happens to croissant volume on a rainy weekday. She knows which seasonal items tend to run out and which tend to linger. That knowledge is real, it is specific, and it produces production decisions that are usually pretty close to right.
The problem is that it does not transfer.
When a second location opens in a different neighborhood, the baker’s read on Location One tells her almost nothing about Location Two. The customer mix is different. The traffic patterns are different. The competitive environment is different. What she has is a starting guess and the expectation that she will calibrate over time.
That calibration process has a cost. In the months it takes a new location to develop its own patterns, the production decisions default to conservative. Make a little extra. Better to waste a bit than to run out. That logic is understandable at the individual location level. Across three locations, then five, then eight, the accumulated cost of conservative production across all of them becomes a structural drag on the operation.
A commissary model intensifies this further. When a central kitchen is producing for multiple retail locations, the production decision is no longer location-level. It is system-level. An error in commissary production does not create waste at one location. It creates waste at all of them simultaneously.
What Changes When You Have a Number
The shift that happens when a production team has a demand-based daily target rather than a safety-margin estimate is not primarily about accuracy, though accuracy matters. It is about what the team is responding to.
Without a target, the question the baker answers every morning is: how much should I make to be safe? The answer to that question has a systematic upward bias, because the cost of under-producing is more visible than the cost of over-producing.
With a target, the question changes: does this number look right for today, given what I know about this location? That is a different kind of reasoning. It invites the baker’s expertise in rather than bypassing it. When the target says 50 almond croissants on a typical Tuesday and her instinct says the same, she has confirmation. When the target and her instinct diverge, the divergence itself is useful information. Why might today be different from what the model expects? Is there a local event? Has the neighborhood changed? Is there a reason to trust her read over the data or the other way around?
That conversation between data and judgment is where good production decisions actually come from. Neither the human estimate nor the model forecast alone is as reliable as both together.
Setting the Right Waste-Stockout Threshold
One calibration question every fresh-concept bakery operator has to answer explicitly is where they want to sit on the waste-stockout spectrum.
Some operators prioritize never running out. The brand positioning and customer experience demands a full case throughout service hours, and a certain amount of end-of-day write-off is acceptable as the cost of that guarantee. Others run lean, accepting occasional stockouts as a signal of freshness and managing customer expectations accordingly. The right answer depends on the brand, the margin structure, and what customers have come to expect.
What matters operationally is that this choice is made deliberately rather than falling out by default. A bakery that ends every day with 15% of production unsold has made an implicit choice about where it sits on that spectrum. The question is whether it is the right choice given the economics, and whether the team is producing to that threshold consistently or producing to anxiety.
When production targets reflect a deliberate waste-stockout calibration, that threshold can be held consistently across locations. The new manager at Location Three is not making a different tradeoff than the veteran at Location One. They are both executing against the same operational intent, supported by data rather than by individual intuition.
That consistency is what scales. Intuition does not. Let’s Talk
