Home / Help Center / pos-app-restaurant / rest-reports / AI Demand Forecast
🧠 rest-reports

AI Demand Forecast

AI Demand Forecast predicts covers, revenue and order counts for each of the next 7 days using a Prophet time-series model trained on your historical order data, with a confidence rating (High/Medium/Low) per day so you know how much to trust each prediction.

📍 Menu path: Reports & Analytics → AI Demand Forecast
👤 Who uses it: All roles with the Reports feature enabled

Overview

Each forecasted day shows predicted Covers, predicted Revenue, and predicted Orders, along with a day-of-week label and a weekend flag.

A Confidence badge (HIGH/MEDIUM/LOW) is computed from the model's prediction interval width relative to the point estimate — a tight interval means HIGH confidence, a wide one means LOW.

The underlying prediction comes from a Prophet forecasting model (the same family of forecasting algorithm used in general time-series demand prediction), trained server-side on your restaurant's historical covers data, then converted here into covers, revenue and order estimates using your recent average revenue-per-cover and pax-per-KOT figures.

Use this to inform prep quantities, staffing levels, and purchasing for the coming week rather than relying purely on gut feel or last week's numbers.

Before You Start

Step-by-Step Guide

1 Review next week's forecast

  1. Open AI Demand Forecast to see the 7-day prediction.
  2. Check each day's Confidence badge alongside its predicted Covers, Revenue and Orders.
  3. Weight HIGH-confidence days more heavily than LOW-confidence ones when planning.

2 Plan staffing and prep around the forecast

  1. Identify predicted weekend or peak days from the forecast.
  2. Cross-check against your usual staffing and prep levels, and adjust for days showing a notably higher or lower predicted covers count than typical.

Every Field & Button, Explained

Field / ButtonWhat it does
Predicted Covers / Revenue / OrdersThe model's point estimate for each of the next 7 days.
Confidence (HIGH/MEDIUM/LOW)Derived from the width of the prediction interval relative to the estimate — narrower interval means higher confidence.
Weekend flagMarks Saturday/Sunday forecasts distinctly, since restaurant demand patterns typically diverge from weekdays.

Tips & Best Practices

Troubleshooting & FAQ

Most days show LOW confidence.
This is expected with limited order history — confidence typically improves as more historical data accumulates for the model to train on.
The forecast seems way off from what actually happened.
Forecasts are statistical estimates, not guarantees — a one-off event (holiday, local event, weather) can cause actuals to diverge from a model trained on typical patterns.