# AI restaurant forecasting

> Forecast the next 7 days of sales, portions to prep and busy hours from your own orders, local weather and the festival calendar. Intelligence plan.

Know what will sell tomorrow, how much to prepare and when the rush will hit, learned from your own orders.

## What you get

- **Sales for the next 7 days,** with a likely range, set against what actually happened.
- **Prep quantities:** how many portions of each item to prepare tomorrow, with a small buffer.
- **Busy hours:** the periods you are likely to peak, so staff and prep are ready.
- **Plain-language suggestions,** such as items guests often buy together.

## How it works, in short

DineMate learns from your own order history and adds local weather and the Indian festival calendar. A gradient-boosted model is blended with a simple same-weekday average, and the mix is chosen by testing on recent days the model has not seen. If the model does not beat the plain average for your restaurant, DineMate leans on the average. Models retrain every day.

## What it needs and what it cannot do

You need at least 28 days of order history to start. Forecasts are estimates with a range, not guarantees, and they are only as good as the history behind them. A new restaurant with little data will see wider ranges.

## Why it is on its own plan

Forecasting is the Intelligence plan. See how the maths works out on the [pricing section](/#pricing).

## Common questions

### How much history does forecasting need?
At least 28 days of orders. More history generally narrows the range.

### Are the forecasts guaranteed?
No. They are estimates with a likely range, tested on recent unseen days, and they improve as your data grows.