Knowing you'll do 200 covers on Saturday is useful. Knowing that 60 of those covers will order the salmon and you're currently three portions short is the number that actually prevents a problem.
From revenue forecast to item-level prediction
Demand prediction takes the same underlying forecasting engine and applies it at the menu-item and ingredient level. Not just "how busy will we be" but "what will people order."
Where this connects to inventory
Item-level demand prediction is what makes NUA's automated reorder drafts specific rather than generic. A purchase order isn't just "more of everything," it's shaped by which dishes are actually predicted to move.
Reducing both waste and stockouts
The two failure modes of manual ordering are opposite and both costly: over-ordering perishables that get thrown out, or under-ordering a popular dish and having to 86 it mid-service. Item-level prediction is aimed directly at shrinking both.
Prediction accuracy improves with more data, a venue's first few weeks on NUA are the baseline the system calibrates against, and forecasts sharpen from there.
What this looks like day to day
A kitchen manager doing prep planning sees which dishes are predicted to be in higher demand for that specific day, not a static "popular items" list, but a live prediction shaped by the day of week, the weather, and current bookings.