August 21, 2017
If you are familiar with the world of Harry Potter, then you know about the sorting hat— the enchanted hat that determines every Hogwarts student’s fate by placing them in one of four “houses” based on their traits. Within moments of being placed on a head, it analyzes the wearer’s character, down to their deepest desires, and uses its findings to declare them a member of Gryffindor, Ravenclaw, Hufflepuff or Slytherin. Imagine how powerful such a tool would be for marketers: if you could get an instant read on your customers based on their history, current context, and aspirations, and boil all of that complexity down into something simple: a profile of their defining characteristics, that you could use to make unique decisions for them. Our abilities would be transformed!
Of course, we do something sort of like this already. Marketers have used personas for decades as a way to distill the deep complexity of an audience into abstracted profiles. We invent a Debbie and a Benjamin, give them basic demographic traits, and use them to represent our customers so that we can picture the people we’re designing for. It’s a great way to drive the creative process--but it can be a blunt tool. Your real life customers don’t likely include an exact Debbie or Benjamin, and when we design too narrowly for imagined people, we can end up making assumptions about our audience that don’t match reality.
In contrast, marketers now have the capability to use artificial intelligence: AI can help us process an unprecedented amount of customer data and act on it in real time. We don’t need to use personas to represent our multitude of unique clients because each person can be defined by their unique data fingerprint--truly individualized. Machine learning brings a level of precision and scale that personas could never approach.
But that’s missing something, too: if AI can parse quantities of information that we can’t even fathom, how can we learn from its hidden process to improve the creative that we design for our customers? If we were to try to extract what the machine has “learned” it would just look like a bunch of meaningless coefficients and variables to us. It may do its job well, but does it help us do ours better? It doesn’t matter if you have the most powerful data processing tool in the world if it’s a black box.
Ultimately, we need something that combines raw processing power with the intelligibility of the persona. The Monetate Intelligent Personalization Engine is the first machine learning personalization tool that delivers interpretable information back to the marketer. We like to think of it as the Harry Potter sorting hat of personalization solutions.
The sorting hat’s calculations remain mysterious, but it delivers an output that everyone can work with. This matters because people don’t think in data: to understand and discuss what’s going on with the numbers, we need a vocabulary like the Hogwarts houses that is based on deep complexity but is simple enough to discuss with our teams. Tools like the persona have persisted because they give us ideas that can be represented with pictures and words, tossed back and forth, and taped to the wall. The Engine does that too--it delivers interpretable information that we can use to innovate--but instead of coming from stereotypes, the profiles it offers are based on real data from each customer. We think that’s much more powerful, and our customers agree.