For consumer goods manufacturers and retailers, machine learning can be a powerful ally.
Just ask Kroger. The supermarket giant’s sweeping “Restock” initiative is a perfect case study in how to put data science to use. Kroger is doubling down on this initiative and increasing capital expenditures over the next three years to over $3 billion. The “Restock” strategic initiative calls for integrating machine learning with other technology components to deliver a seamless personal experience and optimize space at the store level. Kroger’s plans call for touching 20% to 30% of Kroger’s 2,800+ stores in 2018 and upwards to 75% of in-store categories changing on average 8 feet of shelving per category all while reducing cost of in-store operations.
The subtext? Kroger sees itself as a data-science company. As I listened to the company’s October 11 analyst call, I was struck by the scope of Kroger’s plans. So expansively has it bought into machine learning that its digital efforts sound more akin to Google or Netflix than a Cincinnati-based retailer founded in 1883.
Kroger has been signaling its intention with development of 84.51, the company’s 750-person data-analytics unit. According to the call, Kroger can tie 96% of store transactions back to households, whether through loyalty programs, cash, or debit/credit cards. It studies posts and tweets, gathers location data, and knows whether shoppers’ devices are Android or iOS.
Talk about a wow factor. Kroger’s declaration and trust in data to accelerate store level space optimization and assortment changes make it a leader among grocery retailers in aligning to the shopper-driven retail marketplace. It’s moving to market-based demand forecasts and is using machine learning to improve them over time.
It’s a must-do shift. Not only are CPG shoppers moving fluidly as they purchase from in-store and online shelves, but Target’s announcement of voice-activated shopping should convince the last holdouts that the conversational platform has now joined the ranks of shopping alternatives. The result is that CPG manufacturers must define store level assortments across all three shelves: in-store, online and conversational.
The graphic below shows how. By incorporating store level shopper insight, CPG Brands can move from the current push-driven and sales approach leveraging rules to define assortment, planogram, and supply chain fulfilment to one in which suppliers apply data science and machine learning to shopper-driven demand signals.
Machine learning is no silver bullet. Key to getting good insights? Data and the lens through which you view it. While accessing shopper data continues to be a challenge for CPG companies -- many retailers don’t collect the right data, and those that do don’t always share it. CPG companies often have more avenues to develop their own “first party” data. Brands working with multiple retailers typically possess data on their demand space, including detailed understanding of need states and moments, that only product manufacturers have.
It’s the kind of data that will change how you define store level assortment and roll out new products. Without it, you’re flying blind. When combined with retailers’ data, it can optimize assortments across the physical, digital and conversational shelf and improve in-store space optimization and fulfilment. That’s essential information for brands: If you can predict 10% of your business is moving online next year, why pay promotional dollars for in-store shelf space you don’t need?
What’s the payoff? Business benefits include (1) The ability to personalize at store levels, (2) Improving in-store category growth while aligning media and trade spend to shifting shopper patterns, (3) Reduced shelf reset costs and (4) Faster, more successful new product launches.
What’s more, far from its origins in the lofty world of artificial intelligence and statistical science, machine learning clustering algorithms are available through cloud-based, off-the-shelf solutions from companies such as JDA, Enterra Solutions, SAS and Relex Solutions. Digital shelving capabilities to aid planogram design and store resets include Trax’s Shelf Intelligence Suite and Kantar’s VR Infinity 2.0 efforts in virtual reality planogramming.
Machine learning should address a business problem first, and an application of technology second. It has the potential to achieve transformative change for organizations that are willing to trust its recommendations and to commit to operating differently.
Therein lies the challenge: In the work I’ve done with CPG companies, I’ve regularly observed resistance to accepting data insights. But until CPG companies are willing to believe in data -- like Google, Netflix, and now Kroger – they will fall behind in the face of smarter, more informed competition.
Kroger’s vision is what digital looks like. Will everything the company envisions for the new initiative work out? Probably not. Like Amazon, however, Kroger will be able to study the data for feedback and quickly make changes – and that’s where machine learning becomes its greatest ally.