A.I. is Making Retail Smarter Too
“It’s a funny thing,” said Jay McCarthy, VP of product marketing for Qubit. “We’re ironically using tech to be more human again. It’s exciting.”
The London-based product suite, which executes on big data for B2C brands like Uniqlo, Thomas Cook, and Time Inc, is using machine learning algorithms to translate customer signals into the right product offerings. Qubit clients have “maybe millions of customer records, in some cases,” says McCarthy. “The signal-to-noise ratio is quite low.” Qubit’s aim is to “find the needles in the haystack, and scale interactions with customers.”
What that means in practice is aggregating data from online and POS transactions — “There’s an overlap of brick and mortar,” McCarthy say — loyalty programs, CRM, product returns, and logistics — to develop “a more holistic view of the customer; where they live on and offline.”
But this is no idle inquiry, of course. “So much of what you hear [from vendors],” McCarthy explains, “stops at the notion of ‘We’ll give you something insightful from that.’ Even if you have data science, you can have a hard time putting it into play.” The clear aim, for Qubit, is an uplift in conversion rates for their clients; achieved, for the most part, by fine tuning product offerings on the basis of data-driven customer understanding.
What Qubit calls “opportunity mining,” for example, leverages machine learning to understand why some clusters of customers under-perform, in conversion terms, against a complement cluster (users with similar attributes who are converting). In the case of winter coat sales by one brand, the algorithms “bubbled up” geographic differences between clusters. The significance was not that it was winter in one region, summer in another: It was winter all over. The difference was that the under-performing cluster was located in French-Canadian Quebec, and wasn’t being effectively reached by English-language messaging.
McCarthy readily concedes that Amazon algorithms have led a lot of these kinds of consumer-facing innovations. But in the case of Qubit’s AI-led targeting and re-targeted, “Customers don’t even know it’s happening to them.”
Although Qubit’s service to brands is based on the brands’ own data — “Pretty much everything is first party, from the perspective of our customers” — internal meta-analyses go much further. Here Qubit can use aggregated data from across its customer base to drive generalized insights into effective B2C marketing strategies.
One discovery was that cosmetic changes in digital messages — color schemes, text and image placement — were ineffectual in comparison with more substantive conversion drivers like social recommendations.
Machine learning has “gained in popularity over the last couple of years,” McCarthy says, “but we’ve always had that in our DNA. We see this as a key innovation area.” Where are AI and machine learning headed? Towards the dream of one-to-one interaction, McCarthy says. “This is just the beginning.”