The retail industry is undergoing a big change in terms of inventory. As you know, in many countries, inflation figures were very high in the 90’s, inevitably affecting the retail business. The most critical limitation in this space was – ironically – the limited space available. The economic landscape changed in the 2000’s. These were the years when the executives figured out that piling up is not contributing to the bottom-line, on the contrary, it is a pretty bad idea. The concept of ‘Minimum stock level’ emerged and changed the inventory management approach. After almost 2 decades, supply chain professionals are still chasing the answer to a question, where the bar is continually rising; “How can I make my minimum stock policy, even smarter?”. We will discuss in this and our upcoming posts, what a retailer needs to answer this question with confidence.
The 2 main prerequisites of a minimum inventory management approach are;
- Demand Forecasting System
It is no secret that, the success of an inventory management strategy is highly dependent on estimating the demand with the smallest error margin possible. While data flows from origin to the end, each step of the supply chain interacts with one another. Forecasting the demand is the first step of any production system and contributes to maintaining the stock levels, providing better service to customers, improving capacity utilization and bottom-line. Other than these, workforce, materials and capacity plannings are made based on forecasts. It is almost impossible to plan and manage in the absence of a solid forecast.
Retail demand is highly affected by discounts, campaigns and special day events, along with the seasonality and trend effects. To add another challenge, in retail it is common to see items removed from sale and then put back again, for certain business-related decisions. This fact distorts the time series and diminishes the confidence in the forecasts.
With a glance to retail, we see that some stores are located in city centers, some on university campuses, near stadiums, in the airports or on the coastline – heavily affected by seasonal influences. A store in or near the university campus is directly affected by the academic calendar, as a store near the stadium is influenced by the league table. If the products determined to be affected by the special events or dates are not available in the stock room of the stores, customer demand will not be transformed into revenue. Since the effects are of different origins and nature, using a single method to compute the forecasts is not very realistic, after all. For this reason, we employ a multi-algorithm approach. We develop demand forecasting algorithms and race them to select the champion. We choose the best among the forecasting methods (exponential smoothing, regression, Holt-Winters, etc.) according to the nature of the demand and business.
2. Inventory Control System
The results derived from the demand forecasting system are input to the inventory control system. Thus every night, our system checks the constraints; customer service level, the rolling number to box quantity and stock on hand, decides whether the store needs to place an order on any specific product and if the answer is positive, computes the appropriate order quantity.
The value to carry less inventory grows each day and we believe that ignoring the data mining approach and maintaining the inventory management with human intervention hurts the business.
On our next post, we will detail the multi-algorithm structure in demand forecasting.