Auto replenishment of goods with high turnover and short shelf life using hourly sales data

The researches carried out by the authors in the field of auto replenishment of goods within retail chains showed that one of the most important and difficult issues is auto replenishment of the goods, characterized by short shelf life and high turnover. Such products include, in the case of grocery retailers - milk, bread, some kinds of fruit, etc. To maximize a correct calculation of amount of the goods that should be ordered for delivery, it needs to select or develop the best way of demand forecasting. That is due to the fact that an overvalued forecast may lead to write-off because of expiration date, and an understated forecast way lead to the lack of goods available for sale at certain periods of time. In both cases, this leads to large losses for the company, having regard to the high turnover of these goods. The classic and the most common method of demand forecasting is calculation of average daily sales, with or without consideration of trend and seasonality. However for the above-described items the accuracy of the forecasting decreases because of out of stock at certain periods of the day. This article describes a methodology which makes it possible to improve the method of forecasting using a correction of initial historical data for reducing such kind of error. Moreover, in the article is shown a brand new method for calculating the quantity of the goods to be ordered. It takes into account the hour of goods delivery to the store.