The Center for Data Innovation spoke with Withiel Cole, head of data science at SymphonyAI Retail CPG. SymphonyAI Retail CPG is an AI-powered platform that optimizes supply chain performance, improves category and space planning decisions, and delivers a personalized customer experience for e-commerce and brick-and-mortar retailers. Cole explained how AI powers personalization and improved omnichannel shopping, and the future of AI in the retail industry.
Becca Trate: What is omnichannel shopping, and what role does AI play in it?
Withiel Cole: Omnichannel shopping describes a customer who uses traditional and digital channels to fulfill their shopping needs. The strongest growth area for retailers is now being driven by online channels, delivery, and click-and-collect services (and variants of these). While these channels grow, brick-and-mortar remains the most significant revenue source, and rather than customers switching directly from in-store to online shopping, it is now typical for customers to use a combination of these offerings. For example, a grocery shopper might order online for a monthly shop and then make weekly purchases in-store to top-up their supply.
AI provides similar opportunities in the omnichannel environment to the opportunities it addresses in traditional channels. AI can help retailers identify the right products and accurately detect demand signals. It can also help identify the impact of promotional and pricing strategies, as well as understand and predict demand, and manage inventory. In addition, digital channels have unique capabilities and opportunities that AI is particularly well suited for, such as the identification of substitutions dependent on availability or the personalization of promotional activities. Using AI to identify customers who are candidates for omnichannel engagement is also a new challenge facing retailers.
Trate: Shortages in the supply chain continue to plague retailers and consumers. How can AI effectively address these issues, reduce waste, and improve outcomes?
Cole: AI demand forecasting can drive improved supply chain management. In addition, pattern recognition capabilities can observe and learn relationships between products. Understanding the nature of demand transference and substitution effects through observed customer behaviors and item attribution allows retailers to manage impacts by finding suitable alternatives. Systems that can understand these relationships and link them into planogram data provide actionable opportunities to limit the impact of stock gaps on the shelf. (Editor’s Note: Retailers use planograms to plan and visualize a store’s product layout before items are stocked.
Trate: How can data be used to “deaverage” consumers and make more effective predictions about consumer needs and behaviors?
Cole: There are many ways that retailers and CPGs can better understand customers and go beyond readily apparent trends. Detailed evaluation of customer purchases across and between products reveals hidden shared demand that drive aggregate performance—while a particular brand may account for 10 percent of demand, its strong relationship with an alternative may mean that a higher percent of that demand is shared. This has implications for assortment decisions, helping retailers to ensure that shopper needs are met and changes are more accurately quantified.
Automated granular segmentation approaches can identify specific opportunities with groups of customers that may be hidden from traditional insight perspectives. Identification of these pockets through combinations of AI models can reveal trends and emerging behaviors without the constraints of typical product and geography lenses, like removing the reliance on pre-defined hierarchies and moving to learned attributes and groupings.
Trate: How can information gained from e-commerce shopping improve the brick-and-mortar experience?
Cole: The opportunity to better deliver a combined omnichannel offering means incorporating information to gain a deeper understanding of customer needs. This has a variety of implications and can lead to alternative requirements from the brick-and-mortar offering. For example, grocery retailers may need to focus on the top-up shopper with a smaller basket and on fresh and meat sections, and less on large-pack or bulky provisions. Linking requirements met through e-commerce to a realistic perspective of total customer needs can enable retailers to adjust in-store offerings appropriately. Overall, the information gained from e-commerce allows stores to build a more complete understanding of customers’ requirements and respond accordingly. Additionally, a strong online offering provides the opportunity to develop relationships with and better understand customers whose engagement is impacted by physical factors, such as drive time to store or competitor location.
Trate: What do you see as the future of AI in retail?
Cole: There are many ways AI will impact retail, but largely, there are three types of opportunities. The largest current focus within retail is enhanced accurate predictions to improve tactical and operational decision-making, or what we could view as the “atomic level” of decision-making. For example, improved demand estimates drive more efficient and effective supply chains, and accurate, responsive promotional models ensure this key lever delivers to expectation.
Second, e-commerce and personalization will become key differentiators of retailer success. Traditional personalized activity has focused on adjustments to the existing proposition through the provision or to particular financial benefits. As more customers utilize multiple channels, it will become possible to go beyond adjusting propositions through offers to developing personalized propositions, prices specific to individuals, ranges of products that are available online or in-store, or grouped product promotions that recognize known customer preferences. All of this becomes possible with AI, and by increasing online demand, the benefit of developing these approaches only continues to grow.
Finally, while tactical models drive short-term performance, strategic decisions are key to a retailer’s overall long-term performance. Pricing is an example of this. As a tactical decision, incremental price rises often see minimal shifts in unit demand. For instance, a three-cent price increase may impact some customers, but overall revenues will not likely decrease. However, a combination of price rises begins to impact customers and their overall perception of a retailer, driving longer-term engagement and performance. This is already understood, and many retailers manage pricing through rules and competitor indices vs relying solely on accurate tactical models. AI has pattern recognition capabilities that could understand and account for many factors, as models are trained to understand this level of complexity. AI will be able to help inform strategic decisions on how to manage a retailer’s entire proposition with full linkage to detailed tactical predictions that can understand realistic outcomes.