STOCK INVENTORY CONTROL OPTIMIZATION BASED ON SALES FORECASTING

Partner: K2 atmitec s.r.o.

Field: information technology, logistics

Nowadays, it is necessary for production companies to pay attention to efficient use of storage premises. It is far
too complex for people to monitor all stored products in real time, and that is exactly where information technology using state-of-the-art statistical methods comes into play.

K2 atmitec s.r.o. is the developer and supplier of the K2 information system for assisting in control and optimization of business processes via a complex system based on business data analysis. One of the incremental innovations here is the integration of a stock inventory control system, which is based on a module for sales forecasting developed within an established collaboration with IT4Innovations. This module allows the movement of stored goods in the previous period to be analysed, future sales to be forecasted using statistical methods for time series analysis, and excessive stock inventory to be eliminated. As total sales of several companies were analysed during the module development, a huge amount of data was processed, which led to building and comparing numerous statistical models. To verify the effectiveness of the developed module, an enormous number of computationally intensive simulations of stock development had to be performed using the supercomputer. Within our next collaboration, the supercomputer was used to search for the optimal parameters of models for stock inventory optimization.

Regarding the sales forecasting module, the IT4Innovations team deals with data such as order point, safety stock inventory, and economic order quantity set from sales forecasts based on historical data, where the service level can be adjusted by the user. These predictions are based on statistical methods for time series analysis. Various data types are treated: raw data,
purified data, filtered data, and data with daily, weekly, and monthly aggregation. For all combinations, several
models have been built: linear, quadratic, cubic, and ARIMA. Using the MASE (mean absolute scaled error) comparative criterion, each model is automatically compared, and the best one of them is then recommended for further calculations. Stock inventory control optimization based on sales forecasting allows the decision time to be reduced, warehouse operations to be facilitated, and costs to be reduced as a result of stock inventory optimization.