Automatically reduce peaks in the sales history to lessen the impact on forecasting. “High” will reduce more peaks than “Low” and result in lower forecasts.

Setting this parameter to “High” will ensure that a peak (spike) in sales due to an unusual event, will not cause an unnecessary peak (spike) in the forecast (which will result in over-ordering and possible excess stock).

Because this is a global setting, it will apply to all items in the business. It is therefore important to understand the business as a whole:

If the answer is under-forecasting, set the parameter to “Low”. If the answer is over-forecasting, set the parameter to “High”. When in doubt, set this parameter to “Medium”. This works well with regular selling items and those depicting a linear trend.

A common misconception is that if the majority of items in a business have sporadic demand, this setting needs to be set to “Low” so as not to decrease the forecast.

Imagine having an item that sells once a year. In the month it is expected to sell, the last thing we want to do is have the forecast be lower than it should, which would be the case if we opted for reducing the peak (spike) in its sales by setting Peak Replacement to “High”. Right?

Luckily, the app is smarter than that and will not apply peak replacement if the system recognizes the sales and demand type as being young, slow moving, sporadic or seasonal.

Imagine selling 250 units of an item each month. The sales history for this item depicts 250 units each month. In the month of January, this item sold 750 units.

The Peak Replacement parameter is all about how much you wish to smooth out the peak (spike) in sales that will ultimately cause a peak (spike) to the forecast.

In simplest terms, a setting of “Low” will consider January’s sales history to be 750 units, a setting of “High” will consider January’s sales history to be 250 units and a setting of “Medium” will consider January’s sales history to be 500 units, when generating a suitable forecast.

Obviously, this is a very simplified explanation and does not consider the demand type or any other forecasting complexities.

Let’s take a look at a regular selling item with no clear pattern of seasonality, growth or decline. Its forecast is based on a weighted moving average of the sales history. We expect the forecast to be 250 units each month, based on sales history of 250 units each month.

Due to the spike in sales in January, the forecast will be 290 a month.

With a setting of “Low”, the forecast remains 290 a month.

With a setting of “High”, it’s as though the spike never happened, and the forecast will be 250 a month.

With a setting of “Medium”, the spike will be reduced somewhat and the forecast will be 270 a month.

This parameter can be viewed as a less extreme form of Event Correction.