Before finding a solution to the 1:1 real-time marketing challenge, it’s important to understand precisely which hurdles your organization faces. Do you have insufficient data? Poor quality data? Siloed data? Or is your data fine; you just lack the ability to execute on it, in an orchestrated way, across multiple channels? Zylotech is a leading provider of AI-driven customer analytics, both for B2C and B2B clients. At a DMN roundtable, editor-in-chief Kim Davis led a discussion which focused primarily on the concerns of direct and indirect B2C retailers, and elicited some thoughtful responses from Zylotech’s co-founder and CEO, Abhi Yadav.
Zylotech is a Boston-based self-learning decision engine — which in other words means it applies AI to B2C or B2B data in order to enrich customer profiles, predict purchasing behavior, and trigger relevant 1:1 recommendations. Clients include major tech names like Cisco, Dell, and Oracle, and retailers like Staples, Keurig, and SharkNinja.
The direct B2C experience
Miriam Kendall is CMO of M. Gemi, a retailer (primarily but not exclusively online) of luxury Italian leather shoes — “made the old way.” She’s on a journey which she expects to take her from understanding and marketing effectively to individual customers within certain channels, to providing a frictionless 1:1 experience across all relevant channels. She spoke about the obstacles standing in her way.
“For us, the biggest challenge is getting all the data together into one place, and you’ve got structured data and unstructured data (like mentions, and sentiments, and things). I’m trying to marry that with what the person in the call center is hearing, and then tie it all back to one individual customer, and treat them differently according to what we know about them. It’s very difficult to pull it all together.”
M. Gemi is working mostly with first party data, but is also doing some data appends from third party sources. “For email, from transactional data, we can build audiences. Within Facebook and Google too, we can do that pretty well. Within each platform we do a pretty good job of knowing the customer. What I’m struggling with,” she said, “is pulling all of that stuff together, and really knowing the full story.”
Is it a challenge, for example, to know it’s the same customer when they engage with you in different channels, like on Facebook and through the call center? “For sure, 100 percent, and they’re absolutely in all those places. It’s a problem. When you think about the frequency of communication, we know we need like six or seven touchpoints before someone will convert, so we do want to have multiple touchpoints. At the same time, we’re luxury — we don’t want to overdo it. Finding out what works is tough because we don’t have the necessary integration right now.”
A specific challenge for M. Gemi is that while some customers purchase only once, those that buy a second time go on to purchase with a frequency of 3.8 times per year. “Which is actually amazing,” said Kendall. “What we’re working on is that portion who don’t buy a second time.”
The brand also faced a conundrum when it took its first steps with AI. “One of our most popular shoes happens to be liked by older people; so if we let the algorithm do what the algorithm wants to do, we end up targeting the 65-plus range.” Which is not, according to the CEO, the brand’s sweet spot. “If you continue to do that, and go on feeding the algorithm, you’ll continue to make the shoes these people want, and you’ll continue to define your brand in that way.” It’s important, therefore, also to ensure a feedback loop from the target audience, which is younger. “I used to work at UnderArmor, and the customers were not necessarily awesome athletes, but often moms who wanted to be comfortable. That kind of thing happens frequently.”
What’s in your stack? “So we have an email campaign management tool, we’ve got social listening tools, we’ve got reporting tools, but we are really lacking the integration; bringing everything together and making it actionable. That’s next on the list. And having different places for data? That’s a marketer’s nightmare.”
Is real-time, multi-channel 1:1 realistically in your future? “Definitely, and we’re making strides every day, and I think that’s true for everyone. One of the struggles we’re having is the complexity of doing automation: you can optimize within a campaign, and be pretty certain you have a great campaign, but what other campaigns is that person experiencing? You create these business rules, but how do they mesh with the other business rules you’ve set up, and how do you prioritize the rules? Everybody is testing, and doing more and more, but how do you continue to push forward, at the same time doing it in a way that’s thoughtful, and really understanding what the customer is experiencing across all of the marketing channels. It’s a journey.”
Where does a retail brand start if it wants to address these hurdles? “One of the single most challenging parts,” said Yadav, “is the definition of the customer. If someone is in my loyalty system, maybe that’s the nearest definition of customer, but it’s not always the case. Some brands say, anyone who has bought anything from us in the last two years – that’s my customer. You need to know the definition before you start managing the data.”
Yadav had a real-life example of how difficult the first steps in managing customer data can be. “We had this situation with a brand where we were trying to build a single baseline for their customers.” The brand claimed to have enormous quantities of customer data. “They had millions of cookies, and mobile IDs, and email data.” In fact, they had something like 25 million records. “You wouldn’t believe it, but after we did a whole mash-up of the data, it turns out there were only about 7 million individuals, and once we brought it down to households, maybe about 4 million households.” So in addition to knowing what a customer is, brands need to figure out what Yadav called “this whole crazy ID thing.”
For many of their clients, Zylotech recommends a probabilistic approach to identity matching. “There’s no 100 percent when you do this, but as long as confidence in the matches is more than about 80 percent, we can keep on improving using events data.” For Yadav, “events data” is a term which covers the full range of touch points: a customer arriving at a web site, for example, or opening an email, or contacting the call center. Here’s another real-life example. “If we have a client who wants to run a campaign, and the audience is: visitors to my website in the last 24 hours, who are single moms living in Chicago, with lifetime value of $5,000-plus, and who has bought this model recently. From my data pool of 20 million, I can see there are 2,000 of them. I can run a campaign based on that? I can send email or a coupon.” For Yadav, the process isn’t conceptually more difficult than booking a trip on Expedia. “It’s common sense.”
What can leverage that kind of simple starting point into true 1:1 marketing? AI, for one thing. But kicking the process off is what’s essential, even if early steps are based on limited data. “Unless you start, you’ll never have all the data you want. Zylotech offers a complementary data assessment. You might not have like 5000 fields in a customer record, but we can help you go from 20 percent in richness to about 55 percent, by merging in a lot of third party. This should be enough for you to get started, and over a period of time you can bring in other first party. At least you have a data road map.”
The truth is, brands sometimes don’t realize what they already have. “I’ve seen customers who have a lot of data, but no insight. They’re just spraying and praying. But as we start doing a deep dive, there’s richness they’re overlooking.” For example, where a brand has a lot of transactional data, but doesn’t even know whether the individual involved is male or female – “Just by using name logic, you can bring richness.”
Some input from an indirect marketer
The landscape facing Marc Schwartz, COO and general counsel at the Bob Mackie Design Group, is significantly different than that facing a direct retailer like M. Gemi. Mackie is a high-profile designer of costumes and jewelry — “wearable art” — known for dressing icons like Cher and Carol Channing. The Group goes to market via partners, and in particular via QVC, the TV and online shopping network.
“We’re not doing as much direct to consumer,” said Schwartz, “so for us it’s a lot about consumer sentiment. As we have a significant business relationship with QVC, they’re always gathering data, and they’re able to share some of it with us. I’m interested in who are repeat customers; are they coming back; what are they searching for; did they also buy from another brand? It helps us to understand who the customers are, and we can extrapolate that to other parts of our business.” Schwartz is operating at one remove from the actual retail channel or channels, then? “Right, but although we’re not selling the product, we want to be controlling the conversation.”
Both loyalty and acquisition are on Schwartz’s mind. “I’m looking at the core existing audience, and whether they buy once a year or once a month. But then how do you take an existing customer and use what you know to find new customers? Is it best to go out through Facebook or Instagram (we’ve had success with both) versus buying an email list? It’s one thing to keep nurturing your existing audience; that’s one part of it. But how do I find all these other people out there who don’t know about us yet? Customers who are similar to our existing customers?
Key to Yadav’s outlook is convincing brands that data is at least as important – if not more so – than execution. “1:1 is already happening in some ways, but mostly in the last mile of delivery,” he said. “The need brands face is actually in the bottom” of the data foundation. “The journey should be from bottom to top.” Often, brands take the alternative route, starting with what Yadav calls “the last mile.”
“That’s why we see tons of tools there, tons of new things, and people being very experimental. Tons of start-ups. There’s a lot of maturity in these last-mile delivery systems.” If you want to be customer-centric, though, it requires “three D’s” – not just delivery, but data and decisioning too. “It’s the data part where it requires, not just a platform but a whole team around it, and the decision part needs data science.” AI can help. “There’s the AI aspect in a lot of these delivery systems. It’s interesting that it’s now started evolving in data.”