Increasing supply chain pressures are forcing retailers to reconsider each portion of their inventory strategy. They are trying to find the best way to get the right product to the right location at the right time. Modern omnichannel customer experiences add complexity to this. Enabling store order fulfillment (ship-from-store) or adding the ability to buy-online-pickup-in-store can change the way stores interact with customers and requires changes to the retail allocation strategy to ensure adequate inventory is available.
The process of store inventory allocation can be drastically improved with modern analytical and machine learning techniques, addressing core issues that frustrate merchandisers today.
How Does Retail Allocation Work Today?
Most retail allocation done today is either done manually by merchants or driven based on historical sales data in a merchandising system. These systems use sales data from the previous years, seasons, and local trends to generate a recommended value of inventory to send to each store. Some systems can even rely on user-generated store types or clusters to standardize inventory allocations across stores. Apparel and footwear retailers also must adjust for size variations within regions, as well as color preferences in stores. Ecommerce distribution centers are stocked similarly, but with a broader geographic profile taken into account as they are responsible for reaching more customers.
Modern platforms can also offer pre-allocation. This reduces handling time at the warehouse because product packing can be optimized and there is less storage time needed. Merchants can also determine allocation close to when the inventory arrives.
Where Are the Gaps in Allocation?
While today’s methods cover most scenarios, there are still many gaps that can cause frustration for allocators.
New Store Openings
New store openings provide a challenge for allocators today as they have no historical sales performance to base their new inventory on. If the store is in a new geographical area lacking sales data for similar stores, merchants are forced to manually identify a store from a different geography to model allocation. Alternatively they may use a generic allocation model for any new store. This can limit store performance and lead to lost sales either due to not carrying the right product, or continual out of stocks for a popular item.
Prolonged Store Closures
Prolonged store closures (such as those during COVID), can also disrupt traditional allocation methods. Closures for renovations, global pandemics, or hurricanes can create blank spaces in sales data which break up allocation methods based on this store’s sales alone. Allocators today must manually manipulate allocations or recreate them based on another store to account for the closure. This doesn’t consider the specific store’s past performance and doesn’t account for localized shopping trends that appear after the prolonged closure.
Level of Allocation
Merchants make a crucial decision when considering what level of an item hierarchy to allocate on. Some systems allow you to allocate at the class or category level, while others limit to only the item level. Allocating at a higher level can make the model more resilient to outliers, but possibly makes the allocation less accurate. For apparel and footwear retailers, this becomes even more intricate when including color and size into the mix. Allocation models must account for regional differences in size and color, and if the level of allocation is at the style or item level, this resolution is lost in the model.
New Product Introduction
When introducing new products, merchants must manually build allocations as there is no existing sales data for that item. This can be a time-consuming process, involving extracting and consolidating sales from similar products to estimate how this product will sell. Further, allocators are faced with identifying how to allocate this product at each store, relying on grouping by store geography or store types.
What Can Be Done to Improve Retail Allocation?
With these challenges, there are numerous ways that modern data techniques can make a difference. Introducing new data and adding capabilities through advanced analytics can help better handle each step of the retail allocation process, from demand forecasting to inventory management.
Novel data such as local demographics, or climate and weather can add more factors for an allocation model to consider when determining inventory levels. The local population around each store could drive increased purchasing of specific products vs. a store with a different local demographic. Climate and weather can also better determine allocation strategies. Stores in a colder area likely would sell more warm clothing than a store that is in a warmer area.
Beyond that, severe weather can change purchasing patterns. Home improvement stores like Lowes and Home Depot already consider a “hurricane factor” and stock up on storm-proofing products in stores that are in high-risk areas during hurricane season. Having this additional inventory on hand can help handle pre-storm rushes for products and can also have a positive impact on the local community by helping rebuild after the storm.
Those data based on the store location and sales can be used in clustering models to identify groups of similarly performing stores. Unlike user generated groups, clustering models work without user input to find similarities between stores based on attributes or other data to generate groups that share sometimes imperceptible traits.
These models can be applied to new stores, determining which cluster they fit into, and then allocating based on that. This would allow better opening allocation strategies and limit the need for guesswork with a new opening. For reopening after a store closure, the performance of that store’s cluster could be used to drive changes to the allocation.
Clustering could be applied to products too. By building clusters using attribute information (features, size, or descriptors), retailers could then use those clusters as a part of the allocation model. These clusters can be used for new products as well. A new style could be run through the same clustering model to identify similar products, then allocate based on sales data for items that are already in stores. This can reduce the guesswork with new product introductions.
Faced with the rapid pace of change and the unstable supply chain, retailers must find ways to adapt to the new normal. Thinking strategically and leveraging analytical platforms can help improve inventory management without a large investment in a new platform. Retailers should start with a smaller proof of concept project to prove value in these investments, then grow as those demonstrate results to leadership.
Is your retail brand curious about the benefits that advanced analytics can bring to your organization? Speak to one of our retail experts to see how your allocation process can primed for the future
Subscribe to Clarkston's Insights
Contributions by Dave Foos