No empty shelves

Forecasting methods know tomorrow’s demand today

On some days there is just metal on the shelves. A few days later the cash counter rings briskly. Department managers, who can account for each item on their product range weeks in advance, are faced with neither over stocking nor unavailability of stock.

Their customers are satisfied because the shelves are stocked to the optimum. They have no need to maintain surplus inventory for unexpectedly high demand periods and can use the storage space for a greater variety of products instead.

Today, modern forecasting systems are capable of informing months in advance and with accuracy what, when, where and the quantity that will move across the store counter. This is possible even though the demand varies for days of the week, seasons, promotions, price fluctuations and for so many other reasons. “We analyse data from the past with a view to the future,” explains Dr Andreas von Beringe, CEO of SAF AG in Taegerwilen, Switzerland. The resulting sales data builds the foundation for a forecast of every single product – day after day, location by location and for years ahead. Based on this information, the SAF software determines the future requirements for many thousand products. The enterprise is based on mathematical and statistical methods.

The consumer decides

Through this method it is the purchasing pattern of the consumer and not the manufacturer that determines the flow of goods in the logistics chain. The supply chain is transformed into a demand chain – the demand drives the process of value addition. The result is a system of ordering that is strongly linked to the actual customer requirement. By repositioning to the demand chain, dealers kill two birds with one stone: on the one hand they reduce warehouse costs and at the same time increase product availability in the store.

Peak calendar sale dates like Valentine’s, Easter and Christmas as well as promotion dates and special offers build the basis for the forecasting of future demand. Additionally the data for minimum inventory, purchasing costs, purchasing restrictions like ware house costs, date of expiry and packaging sizes are taken into consideration. How formidably enormous the data input that has to be monitored is seen through a case in point of a provision store: with a network of 800 branches with 20,000 items per branch, it adds up to 2.5 billion data entries yearly. This data is accumulated if a branch orders three times in a week. To execute this manually is impossible, not just from the cost aspect but also for reasons of accuracy.

A study backs the result

This is also proved by a study conducted by Metro. The company carried out a three year long CPFR project (Collaborative Planning, Forecasting and Replenishment), in 53 of its Cash and Carry supermarkets. Within its framework the achievement of the SAF forecasting system implemented there as opposed to the traditional methods of ordering were compared. Two test groups concentrated on the procurement of washing agents and detergents, hygiene products and home cleaning agents. One manually adapted the automatically generated purchase proposal. The other ordered as per the purchase recommendations of the SAF system. The participants managed 940 different products from seven purchasers for 53 super markets.

At the core of the study were three criteria: the development of the sales figures, the availability of products and the respective stock volume at the end of the promotion. “The result was unequivocal,” reported von Beringe. “The group implementing the forecasting software strode ahead at all times. During the 36 month long study, the gap between the manually managed trial group increased. In the case of the manually run trial group the availability of all products reduced from 97.6 per cent to 94.8 per cent. Conversely the value with the software implementing group rose from 98.8 per cent to 99.5 per cent. Equally striking was the difference in the product range: it increased with the manually operating group from 24.8 to 37.3 days. In the case of the automatic ordering system the number of days reduced from 22.2 to 18.9 days. In view of the forecasting quality it was observed for all product categories: the automatically generated sales forecasting is so excellent, that a supplementary manual adjustment/application offers no positive contribution.