Sounds like mission impossible, right? Especially when the primary tool is Excel. It’s no wonder planners resort to store categorization (clustering) methods in such a situation. However, it’s worth considering the business consequences of this solution.
One retail network, different store categories
Within clustering, all stores are divided and assigned to several categories, most often based on the turnover they generate. From this, high-, medium-, and low-turnover stores are identified.
In other cases, the sales area is also considered. This especially happens when a brand has several retail formats – for example, large stores in shopping centers, smaller stores on shopping streets, and outlets – which differ in the offered assortment (different products, different collection depth and breadth conditioned by display possibilities), as well as prices (e.g., outlet products with high discounts).
Sometimes other criteria are used during clustering, such as the number of customers visiting the store monthly (data obtained from entrance gate measurements), product turnover rates in the store, or the population of the town where the store is located.
Moreover, there are situations where not only stores but also store sections are categorized. Thus, it is possible that within one store, there are sections marked with different categories.
Assignment to a given category is cyclical, for instance, once per season, and is crucial during allocation, understood as the initial stocking of stores with a new collection of products at the beginning of a new season. For some fashion brands, even at the stage of ordering collections from suppliers, products are assigned to specific stores from a given category. During the season, these categories may change based on product turnover rates, the number of customers visiting the store, or sales volumes.
Consequences of store categorization
What does this mean in practice? Stores assigned to a “better” category during allocation usually receive a more diverse assortment and more products (e.g., a wider size range). Greater product availability, including bestsellers, increases the chances of sales. Subsequently, increased sales mean that during replenishment, these stores will continue to receive the most products, as they will show the greatest potential for selling goods at the first price.
However, clustering stores based on global factors like turnover or area does not necessarily reflect the sales potential of a specific product in a specific store. The risk is that stores classified as high-turnover will remain so because product availability will be high, while stores generating lower turnover (thus assigned to a lower category) will not reveal their potential due to lower product availability. Thus, the principle of a self-fulfilling prophecy will operate.
How do we do it at Occubee?
Avoiding these pitfalls is possible by forecasting sales at the level of a specific product in a specific store, based on historical sales and in conjunction with the historical availability of that product. This is exactly how Occubee operates.
The system enables a comprehensive yet granular analysis of data, providing information on the sales potential for a single SKU in a specific store. It allows identifying local demand and stocking stores accordingly. The platform calculates sales forecasts using artificial intelligence and then generates store demands and picking orders, providing experts with valuable information that facilitates and speeds up business decision-making.
Returning to the introductory example, making a million decisions daily for product-store combinations is impossible for a human. However, it poses no challenge for a system that can make these decisions fully automatically in its algorithm or enable the acceptance of selected results by an expert, leaving control in their hands.
