In the field of inventory management, a stock-keeping unit or SKU refers to a specific item stored to a specific location. The SKU is intended as the most disaggregated level when dealing with inventory. All units stored in the same SKU are supposed to be indistinguishable. Introducing the notion of SKU simplifies most inventory control operations. SKU are sometimes used to refer to intangible items such as warranties, however, in this article we focus on SKUs that have a tangible counterpart.
SKUs vs. products
Unlike a product, a SKU is bound to a particular location. For example for a given product (ex: a book as identified by its ISBN barcode), a retail network can have as many SKUs as there are locations where the book can be stored; typically one SKU per store and per warehouse.
It is also possible to have multiple SKUs for the same item within a single store. Even if the majority of the items are stored at a single location within the store, some items may be sold at several locations, hence generating extra SKUs.
Also a product may have many variants based on attributes such as size, color or conditioning. Thus, even when considering a single location, a single product can be associated to multiple SKUs.
SKUs matter because they represent the most fine-grained level, hence the most desirable level for inventory optimization.
Stock on hand
Each SKU is associated with its stock on hand which represents the number of units readily available for consumption (or picking) at the SKU location. The stock on order represents the amount of inventory in transit that is already due to replenish the SKU in the future.
The vast majority of goods being sold nowadays are packaged lead to strictly integral values for the stock on hand. However, more traditional segments such as fruits or vegetables are still frequently sold by the weight.
As units associated to a SKU are assumed indistinguishable, at any given time, the state of an SKU is defined by its stock on hand value. However, in practice, the stock on hand is not accounted for through direct measurements, at least not systematically. Indeed, only inventory movements are accounted for, and the measure of the stock on hand is typically inferred from the history of movements.
The process of inventorying consists of performing direct inventory measurements, i.e. to count stock on hand values for SKUs. However, this process is intended as a correcting measure to improve the inventory accuracy, and it is performed relatively infrequently because of the manpower costs involved.
Many classic inventory optimization methods poorly handle intermittent demand and low volume sales. However, as SKUs represent the most disaggregated inventory level, SKUs are typically strongly impacted by low volumes. As a result, those methods tend to shift the inventory analysis from the SKU level to the product level because the demand aggregated at the product level is stronger. Similarly, top-down approach where total demand is analyzed first before splitting the results down the hierarchy are also used to mitigate the limitation of classic methods when dealing with intermittent demand.
Yet, we strongly advise not use such methods, because significant errors are typically introduced when the results are disaggregated back to the SKU levels. Modern methods such as quantile forecasting are well-suited to handle even the very low demand levels observed at the SKU level and should be used instead.