Breaking the range paradigm
Bringing together range and space strategy with data science and AI
- Today, retailers are making key decisions about range and space strategies in isolation from each other
- To obtain maximum benefit from physical shelf space, these strategies need to be part of a single end-to-end optimisation
- Machine learning and advanced data analytics have made this kind of all-encompassing decision engine possible for the first time
With growing competition from online players, retailers with physical stores are under pressure to squeeze as much value as possible from every inch of their retail space. Given each store serves different community needs and caters to customers with unique preferences, how can retailers create a differentiated range that builds and maintains customer loyalty, captures a high share of wallet, and drives sustainable returns on their bricks and mortar?
The approach taken by most major retailers with advanced ranging processes is to conduct periodical reviews whereby they analyse the needs of their customers and segment their network into similarly performing stores to generate tailored range plans.
Buyers are responsible for making decisions about which products to keep, which to delete, and which new products to trial in these store segments. Once the range has been determined, attempts are made to optimally distribute it across the retailer’s stores, and a Tetris-like game of fitting products into limited shelf space ensues.
In its various forms, this approach has yielded some reasonable results. Most of us would agree that our local supermarkets offer a wide choice of products and meet most of our grocery needs. In Australia, for example, supermarket and grocery store expenditure accounts for around 60% of total food retailing sales (including out-of-home and takeaway spend).
However, those who are focused on fine-tuning this approach will reach a stage where major efficiency gains have been captured and only incremental improvements are possible.
To gain a competitive advantage in today’s retail environment, decision-makers need to look beyond the paradigm whereby range decisions are made in isolation from space decisions and take an approach that acknowledges that the two dimensions are intrinsically linked.
For example, buyers understand that the sales or profit return generated by a product is not only dependent on the amount of space given to that product, but also on the amount of space given to other products on the shelf – and especially to substitutable and complementary products. The availability of substitutes or complements impacts upon a given product’s demand curve and therefore should be considered when making a ranging decision.
In this multi-dimensional decision space, making range decisions in isolation from space decisions is akin to picking out furniture without having measured up the room first, or without having thought about how the individual pieces will fit together. You risk ending up with a result which is unbalanced, or worse, having to throw out some important items that will not fit. Even if a workable solution can be found, it’s likely that an even better result could have been obtained by tailoring the purchase choice to the space available.
In light of this interactive dynamic there is an obvious rationale for a completely new approach to making range and space decisions, but to this point, the feasibility of an integrated approach has been limited by the sheer scale of such a decision engine, required to perform trillions of scenario simulations across range and space combinations.
The old paradigm treated range and space decisions as a series of separate optimisations, with stores and product performance aggregated up to segments and averages to help keep things simple and manageable. Thanks to an exponential increase in data processing power and advancements in machine learning, we are now able to consider the nuanced relationships between each customer’s needs and their demand for products relative to share of shelf space. Under this new paradigm, we are able to realise the potential of combined range and space decisions and move from aggregated segments to a ‘segment of one’ – unlocking the ability to create store ranges that are accurately tailored to local customer needs.
At Quantium, we are at the forefront of exploring these new possibilities with our clients and have seen five-fold improvements[i] in results from machine learning approaches applied to established retail processes.
If your retail operation is ready to demand more from your range and space strategy than just incremental gains, it’s time to harness the opportunity of an integrated decision engine. Whether you are just starting out or advanced on your data journey, we’d be delighted to speak with you about how data science and AI could be transformational for your organisation.
[i] Quantium Personalisation engine delivers 5x more relevant communications, as compared with traditional approaches