What's the difference between each forecast model?

In the Demand Analysis of every item you can set the forecast model. They are calculated as follows:

Linear Trend - a simple linear trend over the months in your sales history. The system will basically draw a straight line through your last <sales history period> of sales history, then extrapolate this trend to make your forecasts.

Rolling average - the system takes the average monthly demand over the past lead time period, and applies that to predict future months. If your lead time is less than a month, it will be the previous month used.

Steady change - the system will model your forecast as either a decrease followed by an increase, or a decrease followed by an increase. Usually this is not needed, since the seasonality is applied automatically by the system, but if you don't have enough sales history, or it is unreliable then this model is a good alternative if you have fluctuations throughout the year.

Increase then level off - This will forecast your sales as a steady increase from your sales history, and then level to a sustainable monthly demand. This is ideal if you have a new product that you expect your customers to take up in increasing amounts, before becoming regular.

Auto - This is the recommended setting. The system will automatically assess how well each of the above models predicts your forecasted sales. Every time you import your new sales orders, the system will compare the previous forecasts with the actuals, select which model(s) produced the most accurate forecasts, then use that to model your future forecasts.