"BSH Household Appliances" implemented the automated solution on demand forecasting using machine algorithms

2019: Automation of forecasting of sales, calculation of drains and planning of promotion actions

On December 3, 2019 the GoodsForecast company reported that the BSH Household Appliances company implemented the additional methods of forecasting of sales, calculation of drains and planning of promotion actions based on the forecast models developed by GoodsForecast company using machine learning technologies (machine learning, ML).

For BSH the automated solution on demand forecasting using machine algorithms which allowed to simplify planning process of goods of category B and S with low sales level was created.

"The automatic forecast saved to us time for solving of tasks, facing planning department specialists. An important factor is that in the company the system of demand forecasting already successfully functioned, and many processes were built. Implement elements of machine learning without the adjusted business process it is useless", 'Pavel Sobolev, the manager of projects of BSH Russia noted'

Also BSH implemented the automated method of calculation of optimum level of a safety stock in warehouses (so-called Safety Time). A system determines this parameter on the basis of volatility of quality of deliveries and predictability of sales. Thanks to such method of calculation of insurance level of drains the company during project implementation reduced stocks in warehouses approximately by 9% and began to manage more precisely turnover and level of service on different groups of goods.

Before implementation of automatic calculation in BSH the system of automatic formation of orders for the plants in which requirements of production daily formed already worked. But it was difficult to control manually parameters of a safety stock for each type of goods.

It was decided to change approach to calculations of a safety stock and to create a system in which users could adjust a key indicator in a semi-automatic mode, modeling the target objectives of service and goods turnover rate. If traditionally the level of a safety stock is calculated by a standard formula Safety Stock, i.e. in pieces, then in case BSH it was replaced with Safety Time (it is calculated in days). It provided several advantages: Safety Time does not require permanent adjustment as it, unlike Safety Stock, does not depend on a season, it is easier to check it, and it can be average for similar goods.

"Before this parameter was determined on the basis of expert evaluations, without knowing precisely as to prove it. Intuitively understood that on these or those goods it is necessary to put a certain number of days and why and in what results it will result, it was not always clear", 'Pavel Sobolev, the manager of projects of BSH Russia noted'

The analysis of "Big Data" (Big Data) also helped BSH to reduce time for planning and assessment of promotion actions – on average by 20-30 minutes on everyone: the mathematical algorithms developed by GoodsForecast allow to calculate automatically at a stage of approval of a promotion action the basic level of sales cleaned from effect promo. The effect of a promotion action became possible to be estimated as a difference between basic level and the actual sales level.

BSH will continue to apply methods of machine learning.