April 8, 2020
Feng Jia, VP of Data Science
Pareto principle, also known as the 80/20 rule, is an axiom of business management that about 80 percent of revenue comes from 20 percent of customers. As we are confronting one of the greatest health threats of our generation and its resulting economic turbulence, many companies have had to resort to decisions that could have a profound impact on all their customers. Instead of applying blanket treatment/offers to all customers, Pareto principle could be valuable because it could devise differential treatment/strategy to customers who are deemed the most valuable to the businesses.
One example is with one of our online grocery delivery clients. With 90% of Americans under stay-at-home orders, it creates a business boom for them. However, too much demand brought chaos to their online delivery scheduling. Customers were annoyed and bad publicity stemmed through social media. This is a perfect situation where we can have expedited delivery programs designated to the top ~20% of the most valuable customers. Keeping them happy will ensure our client’s business to generate continuous revenue even in the post COVID-19 recovery phase and beyond.
Another example is related to one of our hospitality industry clients. Due to COVID-19, all their customers’ existing resort reservations need to be rescheduled. Instead of opting to the first come first serve approach, they could adapt to an approach where priority booking is designed to the ~20% of their most valuable customers.
One question arises on how to identify the top 20% most valuable customers. There are empirical approaches and predictive approaches. In an empirical approach, one can base on customers' historical value to rank order customers into quintiles, where the first quartile represents the top 20% most valuable customers. When leveraging customers historical values, a certain time window needs to be applied. Typically, industry uses past 12- or 24-months revenue, margin or similar metrics to approximate customer value dimension.
A customer’s historical value is a good approximation for their future values. However, how should we handle customers who have shorter tenure than 12 or 24 months? Or even better, how do we deal with new customers who have little historical transactions with the business? This is when predictive approaches are needed. Under the term of customer LTV, predictive models have been applied to forecast customers future value. The predictive models can be as straightforward as building regression models that predict customer total value within a time window. On the flip side, there could be more complex LTV solutions, where customers churn probability, cross-sell, and upper-sell probabilities need to be considered.
As businesses experience radical changes, Pareto principle should be considered whenever applicable. Analytics and predictive modeling can help businesses to identify the top 20% most valuable customers that differentiated and enhanced treatments can be applied.
