Customer Recalibration Amidst An Economic Crisis
HOW SHOULD I REBALANCE THE USE OF HISTORICAL VS REAL-TIME DATA?
BY CHUCK DENSINGER – COO
If your business has been operating for more than a few years, chances are you have bucketloads of data that have enabled you to forecast business performance with a reasonable amount of confidence. Models help predict everything from product sales to promotion response to customer churn; they fuel product recommendations and email personalization. But the dramatic impacts of COVID-19 on consumer behaviors and the economy itself have now broken most of those models; past behavior can no longer be counted on to predict the future and should not be trusted to be effective.
Does this mean we abandon the use of data to inform decision making? Not at all. In fact, just as predictive models are broken, so are our heuristics and intuitions. We cannot safely rely upon our human expertise and judgment either, as it, too, is based on past experience that may no longer be relevant. This makes the use of customer data—hard facts we can rely on—more important than ever.
With consumer behavior in flux, we must turn, for now, from prediction to signal detection for help. Signal detection is the use of data to identify a shift in a historical pattern. It relies upon statistical techniques to differentiate meaningful changes in customer behavior from random patterns that are not significant or may be ephemeral.
Customer behavior is inherently complex and difficult to interpret and predict. Their decisions are influenced by many factors, some of which are more stable, and some of which change frequently, and which have varying levels of impact on their behavior. A customer’s lifestage, for example, changes relatively slowly, but significant life events (e.g., graduating from college, getting married, having children) tend to be inflection points with drastic shifts in customer behavior that settle into new patterns. While signal detection may not explain the underlying reasons behavior is changing, it can tell us that a meaningful change has occurred.
Used in this way, signal detection can also point to emerging patterns that indicate customer missions are changing. A “mission,” for our purposes here, is simply defined as an in-the-moment want or need; a goal to be achieved, or job to be done. Whereas there is an underlying human core need for food, a mission would be defined as a person’s search in the aisles of their local grocery store for the ingredients to make a meal. One way to view the role of your business in your customers’ lives is your effectiveness in helping them fulfill the missions for which you are relevant.
Customer missions can range from the functional (e.g., replacing a broken washing machine) to the emotional (e.g., signing up for a new video streaming service to lift one’s spirit). As mentioned above, they may result from a lifestage change, or be stimulated by an external event. At present, an external event of historical proportions is driving dramatic changes in customer missions. Our data may not fully explain the functional and emotional drivers motivating them, but they give us important clues.
Here’s a simple example: While we can reasonably expect basic needs like food, medicine, and shelter to remain relatively constant, the customer’s mission around these needs is shifting. Previously, one might have gone to the store as needed for medical supplies and toilet paper, whereas right now we’re on a preparedness mission and stocking up proactively. Our data will show out-of-pattern spending on toilet paper, hand sanitizer, and flour. On the surface, these might appear unrelated. But as humans, we recognize patterns here: in a pandemic, these are supplies we can’t do without, and we’re stocking up.
If you’re like most businesses, you don’t have a documented inventory of the typical missions of your customers. But studying your customers’ “jobs to be done” can be a powerful way to understand where you are and aren’t meeting their needs effectively. The pandemic is stimulating new customer missions and putting others on hold. If we start first with statistical observation of changes in behavior (signal detection), then run the results through our own lens of mission detection, we can arrive at significant insights.
Here’s an example from long before the current pandemic. In the mid 1990s, Target was introducing full-range grocery via the SuperTarget concept. The first SuperTarget store was opened in Omaha, NE, and merchandising analysts noticed some odd patterns in their data, including a lot of lawn mowers being purchased (yes, Target sold lawn mowers back then). In fact, it shortly became the top lawn mower store in the entire chain. Why would adding food cause a spike in the sales of lawn mowers? Talking to front-line employees (who observe customers directly every day and are an often under-utilized source of insight) yielded the answer: wives brought their husbands to the new store to check it out, and the men promptly wandered to more “male-oriented” departments while their wives shopped for groceries. After digging further into the data, they saw spikes in automotive supplies, tools, and home repair products as well.
Traditional product-oriented analysis would suggest adding inventory in these categories and calling it a day. But a customer mission lens causes one to probe deeper: what missions are these men on? Why wasn’t Target their destination for these missions in the past? Will this new trend be sustained, or will it die off as the novelty of the new store fades? And critically, what could we do now to permanently win these customers over as the long-term destination for these missions?
These opportunities are lurking in your data today. The obvious ones, such as the flour shelves being decimated in every grocery store, require no advanced analytics. We know what’s going on … or do we? Have grocery chains dug into who is buying all that flour (which segments), and, critically, who is re-buying, which indicates they’re actually baking with it and not just stashing it in the pantry? Have they studied what else they’re buying that would indicate how they’re using the flour? Can they guess at which customers are more advanced bakers vs. newcomers to bread-baking, for instance, and thought about how to serve their missions differently? Can they turn these pandemic-induced bakers into life-long bakers? Can bread baking be a gateway to other baking and cooking habits, supported and encouraged by grocery stores? How could partnerships with local chefs and bakeries, whose workers are currently furloughed, enrich the experience? And, critically, how would we effectively sub-segment all those flour buyers, and get the right messages to the right customers, supporting their missions and stimulating new ones?
Thus, signal detection has to include the who element—we have to look for micro-signals with pockets of customers, ask ourselves what they are telling us about those customers’ missions, talk to front-line employees or the customers themselves, and start experimenting with ideas to capitalize on what we’re learning.
All that long-term data you have can come into play, appropriately used. The grocery stores, for instance, can probably tell which customers are more advanced home cooks, or are vegetarian, or are doing a lot of grilling, etc. Even in a pandemic, those past behaviors are relevant to interpretation of the new data. But look more carefully than ever for the signals of change in that behavior. And after the new normal settles in, post-pandemic, which it inevitably will, remember those signal detection skills, continue to ask what missions they suggest. You’re building a skill that will be useful long past the current crisis.