By Matt Wampler, CEO of ClearCOGS
Catering is growing. According to Checkmate’s 2025 catering trends analysis, the US catering market reached $72 billion in 2023 and is projected to grow at 6.2% annually through 2032. For fast-casual and full-service brands with active catering programs, that growth is meaningful revenue.
It is also, without the right operational structure, quietly corrupting your restaurant demand forecasting.
The mechanism is simple and easy to miss. A large catering order flows through the POS. It looks identical in the data to a very good day for in-store traffic. A 200-person office lunch order adds a 40% spike above a typical Tuesday’s baseline. If that data feeds into a historical-average forecasting model without being tagged and excluded from the in-store demand calculation, it inflates the prep estimate for every Tuesday in the next forecast cycle.
The kitchen over-preps. Food costs rise. Nobody immediately connects the cause to the untagged event three weeks ago.
Quantifying the Contamination Effect
Consider a practical model.
A restaurant runs a 4-week rolling average to generate its weekly prep forecast. On one Tuesday, a 200-person catering order represents approximately 40% above the typical Tuesday in-store volume. Averaged into the 4-week rolling baseline, this single event inflates future Tuesday forecasts by roughly 10% of the spike (assuming linear weighting), for the next three to four forecast cycles.
| Catering spike above baseline | Weekly over-prep estimate | 4 weeks at 4 locations | Annualized (52 weeks) |
|---|---|---|---|
| +40% on 1 Tuesday/month | ~+10% on forecasted Tuesdays | ~$1,600 over-prep | ~$20,800/year |
| +40% on 2 Tuesdays/month | ~+20% on forecasted Tuesdays | ~$3,200 over-prep | ~$41,600/year |
| +40% on 4 Tuesdays/month | Baseline significantly distorted | Compounding variance | Material structural problem |
These estimates assume $50 average daily prep cost per over-forecasted unit. The actual numbers depend on average meal cost, food cost percentage, and how aggressively the team preps to the forecast. But the directional logic holds: the more frequently large catering orders flow through the POS without separation, the more the in-store demand baseline drifts upward, and the more chronically the kitchen over-preps against a signal that does not represent what walk-in demand actually looks like.
The Scheduling Problem Compounds It
The baseline contamination problem is primarily a prep planning and food cost issue. But it extends into labor scheduling in ways that create a secondary cost.
When a large catering order arrives two hours before service and the labor model was built without it, the kitchen team is suddenly executing a significantly higher volume than the prep decision accounted for. The team that was scheduled and prepped for a $6,000 service is now handling $9,000 in product out the door. Portion accuracy drops under pressure. Items prepared for in-store demand are partially consumed by the catering rush. The day generates high revenue and a poor actual food cost percentage, and neither of those numbers points clearly to the structural problem.
Separating catering from in-store demand is not just a forecasting improvement. It is a scheduling improvement. When confirmed catering orders are logged as known future demand additions rather than historical signals, the labor model can account for them before the day begins.
The Structural Fix
The correct architecture treats catering and in-store demand as two distinct streams that feed into one unified prep number rather than one blended data source that obscures both.
Known catering orders, once confirmed, should enter the forecasting system immediately. A 200-person order confirmed three days out should be visible to the prep model three days out, not on the morning of service. The daily prep guidance then reflects both streams: here is what we project for in-store demand, and here is the additional load from confirmed catering. The prep number accounts for both.
This requires operational discipline: catering orders need to be logged at confirmation rather than at fulfillment. But the payoff is material. The in-store demand baseline stays clean. The catering contribution is explicit and planned for. And as catering volume grows, the forecasting system becomes more precise rather than less.
The compounding math runs in both directions. Contaminated baselines create compounding over-prep that gets worse as catering grows. Clean catering signals create compounding forecast accuracy that gets better as catering grows. The difference between those two trajectories is an operational policy decision, not a technology limitation.
Sources
- Checkmate. Restaurant Catering Key Statistics and Trends for 2025 and Beyond. 2025. [URL needed — verify before publishing]
