Module: Fast adaptation of sales forecast

Executive summary

The module rapidly captures all changes in sales (both increase or decrease). It then analyzes the last few weeks of sales, and if they significantly differ from the sales forecast based on a long-term history (going over several years), then the forecast for the upcoming period is adjusted.

These changes can – of course – occur for various reasons (competition, weather conditions, etc.) and the algorithm is designed to respond to these influences. The module Fast adaptation is particularly useful for fast moving consumer goods and goods with a high elasticity of sales.

Fast adaptation algorithm adjusts a forecast and it adapts it to sudden sales fluctuations.

The purpose of the module Fast adaptation is to detect a sudden nonstandard market development and adapt your offer to step changes in the demand for your products. The Fast adaptation module adjusts and smoothens the current forecast in such a way as to increase the level of customer service – which could drop under the onslaught of demand growth.

Functional description

Computing fast adaptation of sales forecast

The computation is performed over a single inventory item. The algorithm translates (in a simplified manner) the recent history of regression; this is then a base for forecasting the very near (and short) future.

It is possible to adjust both the duration and periodicity of the history. The periodicity sets whether you use a daily or weekly period for the regression. As a rule, for shorter periods and fast moving consumer goods, daily period is better; for a longer periods and slower moving consumer goods, weekly period is suggested.

The subsequent algorithm compares the forecast computed in the module Monthly forecast (split into days in the module Daily forecasts) with the results computed using the regression. If the result is very different, the trend in sales has changed significantly and – for a short period – the forecast is adapted according to the results brought by the regression method.

It is possible to set sensitivity level, degree of customization and also whether the change should (or should not) affect the safety stock (the module: Safety stock).

The algorithm works with seasonality computed in the module Seasonality: it can thus be estimated whether the difference in sales between the most recent period and the forecast is due to seasonality – and, thus, whether something should be adjusted (or not). For example, it is quite expected that sales in December are extremely high (due to Christmas); for January, on the other hand, the forecast is rather low, compared to highly seasonal December. Hence, it is not advisable to increase the sales forecast for January by the regression method.

The algorithm is automatically disabled, if the monitored period has more days than the number of days with promotion sales.

The algorithm does not work if the given inventory item was changed using the extension: Adjustments in the forecasts.

Explaining the module

The module takes into account a multitude of features that can influence the forecast, hence, the user might need a hint for her/his decisions. These hints (explanations) can be found in the module Detail in the tab Forecast.

The outcomes of the module are two charts, always of a monthly forecast and another of the selected periodicity.

The graphs show progress of the algorithms (from the beginning to the end): from the adjustments for stockouts, promotions, to a seasonal index, and the results of the final regression.

The summary table below the graph shows what parameters were and were not met and whether the algorithm was applied to a particular item.

Extensions

Fast adaptation for a group of products

If a company has a central warehouse and a number of branches, then it is essential to monitor the overall growth of sales of the product at all locations, but to evaluate only the central warehouse. It is because the transactions at the individual branches may not be striking (so the algorithm might not detect it), but on the whole, an increase is there.

Consequently, we can attribute this increase to the central warehouse and make sure that it is sufficiently stocked (from all suppliers). The increase, then, can be split equally among the subordinate warehouses, or it can be divided according to the forecast.

This extension allows one to define a group of products: these then are evaluated together.