By Matt Wampler, CEO of ClearCOGS
At some point in the growth of a restaurant group, the tech stack that felt like a complete solution starts showing its edges.
It usually does not happen at location five or six. The team is still small enough that veteran judgment fills the gaps. The operators know the menu, know the locations, know the demand patterns. The prep decision is in capable hands even without a formal system.
It starts to matter somewhere between location 15 and location 30, when the organization has more new GMs than experienced ones. When the ratio flips, the gaps that veteran knowledge was covering become visible as costs.
The Problem With Prep Features That Are Not Really Prep Features
Most multi-unit operators who have built out a serious tech stack have a POS, an inventory management platform, and an accounting system. That combination handles the finances, tracks what was ordered and received, and records what was sold. It does not tell the kitchen team how much to prep tomorrow.
This is the prep-shaped hole. And it is bigger than it looks.
Many inventory platforms have added prep modules. The feature shows up on the product page. It sounds like it solves the problem. Then operators go to use it and find that it requires manual input to the point of negating most of its value. The system can calculate theoretical usage from recipes, but someone has to maintain those recipes precisely, update them when portions change, and make sure the POS mappings are current. When those things are done well, the tool can be useful. When they are not, and they often are not across a growing portfolio, the tool surfaces numbers that the kitchen team cannot trust.
The result is that many operators with a full tech stack still have GMs making prep decisions the same way they did at location one. By feel. Based on how last week went. With a buffer because nobody wants to run out.
What Over-Prep Actually Costs Across a Portfolio
The instinct to over-prep is not irrational. An 86 during service is visible, embarrassing, and costs guest trust. Waste is invisible and gets absorbed into food cost variance without anyone attributing it to a specific decision. That asymmetry shapes behavior at the GM level in entirely predictable ways.
According to a 2026 analysis from WifiTalents, pre-consumer food waste accounts for 58% of all restaurant food waste, and over-production drives 45% of total waste in professional kitchens. For a fast-casual brand with high-volume, short-shelf-life proteins, that is not a rounding error. That is a structural margin problem.
The labor dimension compounds it further. The prep labor that goes into portions that never sell is not recoverable. Prepping 200 portions when demand will support 110 is not just a food cost problem. It is a labor allocation problem. That time could have been spent elsewhere in the kitchen.
Why Judgment Does Not Standardize Across 50 Locations
Good judgment is location-specific. A GM who has worked Tuesday mornings at the same site for three years understands the demand pattern for that specific location. She knows how the season affects traffic, how local events move volume, how the first week of the month differs from the last.
A GM six weeks out of training at location 47 does not have any of that. She has the training manual and the instinct to stay safe. The prep decision she makes is not wrong given what she knows. It is just expensive relative to what accurate demand intelligence would produce.
This is the scaling problem that does not appear in the unit economics model at location five but shows up clearly when the portfolio has 50 sites. The variance between how experienced locations prep and how newer ones prep is real, and it accumulates across every service period at every location.
What a Forecasting Layer Actually Does
Forecasted prep is not a replacement for kitchen judgment. It is a correction mechanism that ensures that judgment is applied from a better starting point.
When a GM sees that the model predicts 100 portions of chicken on a Tuesday and her gut says about the same, she gains confidence. She preps 110 instead of 200. When the model predicts 80 and her gut says 120, she has a conversation with her data before making the call. Both outcomes are improvements over gut-only decision-making at scale.
The forecasting layer reads from the POS and recipe data that already exists in the stack. It does not replace the inventory platform or the accounting system. It sits on top of both, turning the historical data that those systems collect into a daily prep recommendation. The investment operators have already made in their tech stack starts working harder.
Onboarding is lighter than most operators expect. The first prep sheet looks similar to what the kitchen already uses. The difference is in the numbers behind it.
For operators who are 20 or 30 locations in with the gap still open, the cost of waiting is real. Every new location added without closing it adds another team making prep decisions on instinct, and another set of months before that instinct catches up to what the data already knows.
