Grocery retailers in Germany curate and offer articles from over 10,000 different manufacturers. This number is continuing to increase as customers keep demanding new brands and products. Over the last ten years, we have seen assortment breadth increase by up to 20%. This is reflected in growing average floor space per grocery outlet of around 7% across segments.

Today’s consumers embrace choice and rather buy from a portfolio of stores than being loyal to a specific outlet. On average, consumers visit 2.9 different stores to cover their weekly needs. But is this what they really want? We know that at the same time, consumers prefer grocery retailers that are easily accessible, convenient and offer a one-stop-shopping experience. Consequently, consumers will turn to those retailers that can best address these requirements at one single place. Therefore, category management and localizing assortments are becoming more and more important again.

The size of the prize is considerable: Consumer-centric assortments are THE key lever to sustainably grow net sales, profitability, and increase loyalty. Consequently, assortment optimization is experiencing a renaissance. The most beaten truism of retail – the right product at the right time and the right place with the right price – proves its applicability once again. This evolution is fueled by the proliferation of data and advanced analytics and AI. While assortment optimization approaches have significantly evolved, there is still large unrealized potential when it comes to truly mastering consumer-centric and hyper-localized assortments.

Today there is already a wide array of solutions in the market offering assortment optimization functionality to automate category management tactics. Mostly, these rely on attribute analytics-based approaches and restrictive, parametric models which optimize expected net sales and margin, while accounting for cannibalization effects and connecting with planogramming applications for swift implementation. These types of solutions are flexible in terms of location level – i.e. analytics can be performed on a chain, store cluster or even individual store level. However, very often, they do not support decision making regarding the right level of differentiation vs. commonality.

The optimization problem of striking the right balance between hyper-localized and common assortments remains at the discretion of Category Managers. They need to weigh the benefits of differentiation vs. increasing complexity costs and the potential impact on sourcing, supply chain and category management. Often, a cluster-level localization may be preferrable to a store-level localization and allow to strike a reasonable balance. Leading models and solutions rely on an accurate application of a reliable planning foundation (historical performance and reliable forecasts), consideration of strategic positioning (balance between historically proven high-performing SKUs and strategic SKUs), understanding of assortment dynamics (controlling for availability or absence of alternative options) and accurate representation of operational cost structures.

Even though assortment planning has evolved from a manual, gut-feel exercise to a data-driven analytical process over the past 40 years, most productive solutions today still rely on considerable simplifications.

To stay on top of an evolving customer base, three enhancements are necessary to sharpen retailers’ understanding of consumers, leading to high-performance assortments:

1. Use of advanced modeling techniques based on personalized data
2. Integration of external data contextualizing consumer decision-making
3. Reflection of cost and requirements structures (e.g., supply chain operations.)

Implementation of state-of-the-art assortment optimization solutions is not an overnight process. Analytical and business processes need to evolve over time and in sync with ongoing operations. Furthermore, the change aspect cannot be overestimated in its importance for success: Category Managers’ acceptance and adoption of data-driven decision making for assortment optimization is fundamental for AI to unfold its impact on value. This translates to a multi-year transformation effort. Generally, an evolutionary development of existing solutions to the target picture has proven to be very successful.