Someone recently asked a friend of mine, the director of Audience Strategies Ltd., David Boyle, “What is one marketing idea that very few people agree with you on?” His response was quick: Clustering is everything. He has a deep-seated belief that just about every piece of analysis would be more insightful if clustering is used.
When I asked him why he felt this way, he described what he considers a transformative experience in his career. He once worked for a retailer that made a lot of assumptions about its customer base, including that most were one-and-done customers, meaning they came into the store, purchased a single big-ticket item, and never came in again. The in-store sales reps and marketing team were sure this behavior represented a large portion of its customers.
Skeptical, David decided to question those assumptions by leveraging sales transaction data to cluster the store’s customer base. He clustered using different variables, such as the number of divisions (or departments, such as menswear versus home goods) they shopped in, spend per division, total spend, and so on. The exercise was enlightening. First and foremost, it turned out that the store had exactly zero clusters of customers who purchased just one big-ticket item only, and that, in turn, meant the brand was needlessly dismissing a large portion of their customers. And while there were indeed customers who purchased a single big-ticket item, they also shopped in other divisions, and visited the store at least five or six times a year.
“Clustering allowed us to bust the myths. More than that, for every customer we now had a view of who they were, which divisions to engage them in, which products to highlight. For me, it was transformative. It opened my eyes to the value of boring old customer data. Once it’s clustered, it’s gold,” David said.
What is clustering exactly, and why is it transformative?
Clustering is similar to customer segmentation, but rather than looking at a single dimension or variable, we look at many. In other words, rather than asking how new customers perform against existing ones, ask how existing customers who meet a specific spending threshold, shop across multiple categories, and visit a store at least once a month differ from those who spend the same amount, but visit a store less frequently and shop in fewer categories?
Clustering will reveal an astonishing level of nuances that allow marketers to identify groups of like-minded people within their customer base. It unlocks profound insights that can’t be seen when looking at customers as a whole.
The deep customer understandings that result from clustering can (and should) inform every task a brand engages in, from product development and merchandising to message development, promotions and targeting. More than that, clustering enables a company to view its business from the lens of its customers, rather than its products. That’s game-changing.
I’ll give you an example. Retail chains tend to look at sales data and ask, how did our divisions, departments, and products perform over the past week? Which brands were the best sellers? Which stores saw the most sales? Can we increase revenue if we ship more of the fast-moving products and brands to the stores with the highest sales?
However, I argue that it’s much more strategic to ask: how did each type of customer perform this past week? How can we engage each customer base so that they visit a store? Which message should we use?
How to cluster at scale
Every brand needs an audience strategy. Who are your existing customers and who are the customers you want to get? This can certainly begin with traditional marketing research, but don’t limit it to that. Dive into a wide array of datasets to discover what your audience cares about, who influences their world views, sites they visit, and so on.
Once you have this data you can begin clustering your audience according to behaviors, and watch the magic unfold. For each cluster ask yourself: what is the next best thing we can do for them? The answers become your engagement roadmap.
Contrast this approach to the marketer’s previous thinking, which says one needs a big research study to get the kinds of rich data on customers needed in order to create meaningful clusters. The revolution here is that even boring old transaction data along with augmented analytics platforms will deliver meaningful clusters. That’s a real revolution, as it shortens the time and costs of doing market research. It’s as if marketers now have an X-ray machine that allows them to look inside their datasets and see incredible richness.
Affinio provides what it calls an “augmented analytics platform” to uncover hidden insights in marketing data.