Most restaurants plan staff and stock from memory: the rota is copied from last week, and the order is “the usual, plus a bit just in case”. A sales forecast built on your own history makes the busy and quiet days visible in advance. This guide explains how AI sales forecasting in Kappino works, from the data it learns from to the decisions it leaves to you.
Step 1: your orders build the history
Every dine-in, takeaway and delivery order you take in Kappino becomes training data. There are no spreadsheets to upload: the history builds itself as you trade. Forecasts improve as that history grows. The peak hours heatmap needs at least four weeks of orders, and covers forecasts rely on the guest counts of dine-in orders.
Step 2: the model trains on your data
You start training whenever you like and follow its progress live. Kappino tests several forecasting models on your recent data and uses the one that performs best for your restaurant. Public holidays, seasonality and day-of-week patterns are built in, so a bank holiday or the summer peak does not catch you out.
Step 3: read the forecast
Everything sits in one Forecasting view, updated from the orders you take:
- Revenue, covers and average order value over the next 7, 30 or 90 days, with a confidence band around every prediction.
- Peak hours heatmap: predicted order volume by day of week and hour, so you can staff the rush instead of guessing it.
- Top demand items: the dishes expected to sell most over the next seven days, useful for prep lists and ordering.
- Scenarios: optimistic, baseline and conservative forecasts side by side.
- Anomaly scanning: a scan of the last 30 days for unusual swings in sales.
- Written insights drawn from the forecasts and anomalies.
How much to trust a forecast
Every forecast carries a badge: high, medium, low or preliminary. It stays preliminary while your history is short, and accuracy is only shown when the model is reliable. The badge tells you how much weight to give a number before you plan around it.
Step 4: decide what to act on
The forecast comes with recommendations on pricing, staffing, reordering and promotions, ranked by expected impact. You apply or dismiss each one. Nothing changes in your restaurant automatically: a recommendation only changes status when you act on it. The forecast can also be exported as CSV.
Putting a forecast to work
- Rotas: use the peak hours heatmap to put people where the rush will be. The AI shift planner builds rotas around forecast demand, availability and labour cost.
- Ordering: top demand items show what to prep and buy for the week. For ingredients, AI Inventory predicts next week’s use from the last 90 days of consumption and suggests how much to buy.
- Promotions: the quiet days in the forecast are the ones where a promotion can make a difference.
- Prices: pricing recommendations sit next to AI smart pricing, where you decide what to apply.
Which plan includes it
AI Sales Forecasting is part of the Professional plan. See pricing for what each plan includes.
