Targeted Serendipity: Thinking Harder About 'Relevance'
Perhaps it's because MediaPost Editor In Chief Joe Mandese and I are the same sort of ink-stained humanists predisposed to suspicion about too much automation. But at the end of our back to back OMMA DDM (Data Driven Marketing) and OMMA RTB shows, there was one panelist observation that kept echoing in our heads: a point about the danger of marketers being too relevant all the time.
On the first panel of brand marketers (see full video here) at OMMA DDM on Wednesday, executives from Hearst, Macy’s, JP Morgan Chase and Starwood Hotels drove home the importance of relevance to the consumer. Macy’s marketer Julie Bernard, however, called out an important exception to that rule. Personalization and customizability can be overdone, she warned. If the brand hones in too tightly and exclusively only on the things for which a user has shown an interest, then “You’re only ever showing me apparel and handbags,” she said. This technique eliminates the important process of discovery.
Who better to understand the value of discovery than a classic retailer? The modern department store is pretty much based on the principle of browsing and happening upon the stuff you didn’t know you wanted or needed. And for women, who still do more shopping for others than most men, getting outside one’s own shopping zone is even more important. “Maybe I want to buy something for my husband once in a while,” Bernard said. How is your personal purchase history going to help the algorithms there?
To be sure, her point has to be couched within the larger value of relevance and what we really mean when we use the term. Macy’s has pioneered a number of projects that used data for one-to-one marketing efforts. The point is not that a marketer shouldn’t leverage data to be as relevant to the consumer as possible, but that relevance too tightly defined and too obsessively pursued feels mechanical and can miss a larger branding goal.
As Julie shared with me later, Macy’s uses advanced modeling to target its physical and email communications to the households most likely to want them. The company especially focused now on using these data points to understand what form of communication to make to the right customer.
Macy’s is also trying to figure out if delivering more pages of relevant content to an interested consumer has a net productive effect. The retailer found in these tests not only that relevance matters, but is a key driver of sales -- and that limiting a mailing to the categories that scored as relevant did not negatively impact overall store sales. But Macy’s is also trying to determine how to encourage cross-shopping.
I asked Julie to reflect on her insight further on what it means to be too doggedly relevant. She tells me, “So, if we see that a customer has only ever purchased red lipsticks, it doesn't mean we should only serve her content and offers around red lipsticks. Certainly, that would be relevant but it wouldn't deliver against our entertainment brand promise to help our customer find their magic through My Macy's.... In our example above, rather than simply serve red lipstick content, we would be sure to serve [other] relevant content, which may be pink lipstick for the seasonal trend, coupled with a fabulous wristlet to carry it in. An over-simplified example, but it gets to the spirit of how we are advancing relevancy for the benefit of our customers.“
So serendipitous discovery is not necessarily “irrelevant” material, as much as content relevant in a way that a single customer profile does not reveal. As Julie puts it, “The wonderful thing about analytics and relevancy is that we can infer a customer's preferences and then model the behavior of other customers to understand propensities for interest in other categories, thereby introducing the discovery element for the customer."
I would argue that a certain tedium sets in when we get targeted the same kinds of items from the same source. For example, while I still regard Amazon's recommendation engine as the gold standard for personalized experiences, it has become so well tuned to the patterns I have established over the last dozen or more years of buying that the items it suggests are predictable to the point of invisibility.
I am dwelling on the very human responses to machine-made “personalization” because that is one of the biggest takeaways for me from both drafting the OMMA DDM show and watching the discussions unfold Wednesday. The chief irony of DDM is that these highly technical, complex, algorithmically driven mechanisms behind “big data” actually force marketers to think about their customers in more rounded human terms.
That is the key message I got from the brand marketers on that first panel. I compare the massive number of new inputs available to a brand about their customer to a pointillist painting. There are so many dots, each of which alone is without meaning, but when ordered correctly and in such huge numbers create a recognizable impression of a person in her context. It doesn’t seem to me coincidental that the cold, technical terminology around “data,” “modeling” and “algorithms” emerges at the same time as much more poetic references to a consumer “journey” or “path.” The technology helps render a more poetic, lyrical understanding of the consumer.
But that also means that the machines that interface with humans need to seem more human. For instance, my personalized Amazon page now confronts me with an uncomfortable mirror of only a slice of my reading habits. That is not “personalized” in the richer sense, so much as deftly targeted.
At some point I wonder if data-driven marketing has to start building into the system either randomness, or a smarter kind of tangential relevance or targeted serendipity.