For restaurant brands that operate a central kitchen or commissary model, the daily logistics of getting the right product to the right store in the right quantity is one of the most labor-intensive processes in the entire operation. And in most cases, it is also one of the least automated.
The workflow looks something like this: each store reviews what it needs for the week, submits an order to the commissary, someone at the commissary tallies up orders across all locations, production is planned based on those tallied numbers, and product is distributed. The process works. It has worked for years. But it depends on manual aggregation, manual communication, and manual adjustment at every step.
When any link in that chain breaks, whether it is a store manager who forgets to submit an order, a commissary team that misreads a tally, or a production run that does not account for a sudden demand shift at one location, the entire system feels it.
The Tally Problem
In a commissary model, the central kitchen needs to know the total demand across all stores before it can plan production. This means someone has to collect individual store orders, aggregate them into a production summary, and communicate that to the baking or cooking team.
For brands with three or four locations, this is manageable. For brands with six, ten, or twenty locations, each with different demand patterns and different order cadences, the aggregation process becomes a significant time sink. The person doing the tallying is usually a manager or owner who is also responsible for a dozen other things. The process is repeated weekly or even daily, and errors compound.
The underlying issue is that store-level demand and commissary-level production planning are disconnected systems. The stores know what they need based on their own experience, but that knowledge lives in spreadsheets, text messages, or verbal requests. The commissary knows what it can produce, but it does not have real-time visibility into what each store will actually sell. The gap between the two is filled by human judgment and manual arithmetic.
When Par Levels Hit a Ceiling
Many commissary-based brands start with par-level forecasting at the store level. Each store has a set number for each item: keep 10 bags of this, 5 trays of that. When inventory drops below par, the store orders more from the commissary.
This approach provides a baseline, but it has significant limitations. Pars are static. They do not adjust for a Tuesday versus a Saturday. They do not account for a holiday weekend, a new promotional item, or a seasonal shift in demand. And they tell the store what to have on the shelf, but they do not tell the commissary what to produce until the order comes in.
The next level of sophistication is demand-driven forecasting: using historical sales data at each store to predict what will be needed before the store even places an order. When a forecasting system can project that Store A will sell 45 egg tarts on Wednesday and Store B will sell 28, the commissary can begin production planning before a single order is submitted.
This shift from reactive ordering to proactive production planning is where the most significant efficiency gains live. It reduces the manual tally process, shortens lead times, and gives the commissary team the ability to plan production runs based on aggregate demand across all locations rather than waiting for individual orders to trickle in.
Closing the Loop with Waste and On-Hand Data
One of the most common data gaps in commissary operations is the feedback loop between what was produced, what was sold, and what was wasted. Without this loop, the forecasting system is operating with an incomplete picture.
If a store receives 50 pineapple buns from the commissary, sells 35, and wastes 15, that waste needs to be visible in the data. Otherwise, the next forecast treats the store as having consumed all 50, and production creeps upward over time, creating a cycle of overproduction that nobody can trace back to its source.
Similarly, if a store is supposed to produce 14 of an item but only puts out 5 and holds the rest for the next day, the sales data shows a spike the following day that does not reflect actual demand. The forecasting model interprets this as a real demand increase and adjusts future projections upward. These operational inconsistencies at the store level create noise in the data that, if unaddressed, degrades forecast accuracy over time.
The solution is to capture waste and on-hand inventory data at the store level and feed it back into the forecasting system. This does not need to be complicated. It can be as simple as entering waste counts into the POS at the end of the night or logging on-hand inventory before closing. The key is that this data exists, is captured consistently, and flows back to inform both the store-level forecast and the commissary production plan.
From Store Forecasts to Ingredient-Level Ordering
The natural evolution of commissary forecasting goes beyond finished products. If the system knows how many egg tarts each store will sell, and it knows the recipe for egg tarts, it can calculate how much flour, how many eggs, and how much custard the commissary needs to produce the total volume across all locations.
This transforms ordering from a store-by-store guessing game into a centralized, ingredient-level procurement plan. The commissary can go to its suppliers with precise volume projections for the next week, the next month, or the next quarter. That visibility creates leverage in purchasing negotiations and reduces the risk of both over-ordering and emergency reorders.
For brands planning to scale, this kind of ingredient-level forecasting is not a luxury. It is the infrastructure that makes growth sustainable. Adding a new location to the commissary network should mean adding one more set of demand signals to the model, not adding one more manual tally to someone’s weekly workload.
Your commissary should know what to produce before your stores know what to order. That is the gap forecasting closes.
