Customer Recalibration Amidst An Economic Crisis
HOW WILL I KNOW WHEN MY CUSTOMER PORTFOLIO IS STABILIZING?
BY JIM SAWYER – CHIEF SCIENTIST
If you are like most companies, over the past several months you have been working tirelessly to adjust to the new economic climate—changing how you do business, starting to reopen operations or planning for reopening, and modifying your expectations of “normal” customer behavior. The term “new normal” has become a common phrase to describe a general expectation that people will not return to the exact same behavioral patterns as before the pandemic. While the effect of the virus still has no end in sight, you still have to continue making informed, long-term decisions about your business. The question is, what signals within your data should you trust as more than merely temporary blips on the radar?
For example, panic buying at the outset of the pandemic disrupted our supply chain, so much so that toilet paper manufacturers are still working to level out the balance of supply and demand again. And just try to find disinfectant wipes at your local grocery store. However, we shouldn’t expect the true need for toilet paper and wipes to forever stay at the crazy levels of these early days—the pandemic didn’t transform us all into a nation of hoarders.
Conversely, we have also seen a massive shift in some aspects of consumer behavior since the crisis began, especially for those of us who are now working from home (e.g. telemedicine, online shopping with curbside pickup, Zoom meetings)—and these trends show no signs of slowing. You have likely already placed a heavier emphasis on your longer-term digital strategy to stay on top of this move. What other signals are you seeing that warrant your attention, and might indicate the beginning of a stabilization period?
Actually, “stabilization” is a bit of a misnomer—for most industries, it will take quite some time to learn what the “new normal” means for your organization. But this does not mean you should wait to begin your analysis. If you were lucky enough to stay open during this crisis, or even if you’re just beginning to re-emerge in recent weeks, you have data about your customers and their current purchasing patterns. And you should be leveraging this asset to make smart, data-driven bets on the future direction of your organization.
First of all, those year-over-year same-store comps you’ve been using out of habit for years can be thrown right out the window for the next year or so now that the world has been disrupted. They’re only useful for the shock value at this point. Instead, as we outlined in one of our previous articles, “How Are My Core Customer KPIs Performing,” there are three core KPIs you should evaluate, namely: number of customers, transactions per customer, and spend per transaction. At a bare minimum, you should be capturing these KPIs on an ongoing basis and analyzing trends on a week-over-week and month-over-month basis. These metrics are often used by start-ups with less historical data, relying on more near-term sales patterns to evaluate the trajectory of the business. In effect, COVID-19 has knocked many companies back into start-up mindsets, such that a refocusing on these simpler metrics makes solid business sense.
But looking at these trends in aggregate isn’t nearly good enough—your customers are not all the same as each other. You should be looking for patterns within different dimensions that are relevant for your business and industry, including customer segments (if you have them), channel, geography/region, and product category. Different groups of customers may be interacting with your brand in different ways, and you want to learn from this behavior to guide your decision-making.
Keep in mind that for this level of customer analytics, the goal is not to challenge yourself to identify patterns that indicate a larger-scale, economic turnaround. Tracking macro-economic trends via the right external data sources, as we’ve also written about in “What External Data Should I Be Looking At,” can be helpful for that endeavor. But you don’t need to be a trained economist to evaluate when your business is emerging from the crisis. In this case, you’re hunting for micro-trends within the data you are collecting about your unique set of customers.
However, looking at historical activity to monitor customer behavior is just the starting point. Even with just a few weeks or months of data, you can begin leveraging machine learning techniques to predict what you might expect to see in the weeks and months ahead. And you don’t need to buy the fanciest AI solution on the market to do this—this is the realm of time series forecasting, a tried-and-true set of techniques that has been used by statisticians and data analysts for many years.
Time series forecasting, when done correctly for relevant customer groups, can help you decompose certain components of the data—seasonality, noise, and actual trend—that will allow you to make more informed decisions about which patterns are emerging, and which you will need to collect more data for before concluding there is actually a pattern. There are also modeling techniques (note: for any data scientists or stats nerds reading this, look up “Bayesian structural time series”) and supplemental methods that can help you quantify the uncertainty around future predictions of purchasing behavior.
Finally, remember that averages are evil. Even within customer cohorts that are similar in some respect (e.g. region, spend band, channel), individual customer behavior spans a distribution of values. For each KPI or behavioral attribute you are measuring and monitoring, you’ll want to review how the distribution is changing over time. A simple yet effective tool to do this is to capture not just the average, but the “five number summary”—the minimum, 25thpercentile, median, 75th percentile, and maximum values within the range of the variable measurement. Note that when the distribution of a particular variable is “skewed”—when there are a lot of small values such as low-frequency purchasers or a lot of high values like revenue from your best customers—the median is more effective to track over time than the average.
In conclusion, many of the new customer needs we are seeing warrant more than a short-term reaction from companies. In that sense, the “new normal” is less of a state of being that is eventually realized, and more of an evolution over time that you will need to adjust for based on data. The companies who can detect and react to the right signals will emerge from this crisis as victors. This is not survival of the fittest; rather, it’s the survival of the most adaptable.