“Predictive intelligence” has rapidly gained a position of prominence on the ever-expanding list of retail and marketing buzzwords. Behind the hype, however, is a powerful capability: using computer models to predict some aspect of a customer’s future behavior based on their observed past interactions with your brand.
A wide and diverse range of methodologies culled from statistics, machine learning, and artificial intelligence can be deployed to implement these models, and the data science community has developed formal methods for evaluating the accuracy and effectiveness of their predictions. The objective, ultimately, is to identify – ideally at an individual customer level – where limited marketing resources should be invested to maximize return.
‘Next’ Is a Shortsighted View
Currently, most of the action seems to revolve around predicting what the customer is going to do next. What’s the next product category we should recommend that will get them to buy? What should we email them next? When’s the next time they’re going to come into the store or visit our website? What’s the next best action for us to take? These are certainly important questions to ask and an appropriate class of problems that predictive modeling can help inform. But next can be a shortsighted view.
Unless your predictive intelligence practices are grounded in a well-quantified, data-driven understanding of where your customer is at in his or her customer lifecycle — their evolving long-term relationship with your brand — you run the risk of solving many small problems and missing the big ones. Marketing and communications strategies are more effective in the long term if they are developed uniquely for customers at each stage of the customer lifecycle, differentiating at a minimum between new, active, and lapsed customers, and setting strategic direction accordingly for each of these groups.
Strategies to develop new customers should be distinct from strategies to retain active customers and strategies to reactivate lapsed customers. Within each stage and each strategy, predictive modeling can provide a valuable set of tools to help generate insights that guide your decision-making.
Predictive Intelligence and Your Customers
Consider your new customers. Great news — your acquisition strategy is working and you’ve just acquired a brand new customer. It might be tempting to use predictive intelligence to answer the question: Based on what this customer purchased the first time, what’s the most likely item they’ll purchase next? And then you could set up an email campaign to try to convince the customer to purchase that specific item. Predictive models can certainly help with this.
Alternatively, however, you could consider changing the question: Based on what this customer purchased the first time, what is the probability that this customer will become one of my best customers? Which new customers have the highest potential to become my most fervently loyal brand fanatics? By studying what is known about your best customers, and then using predictive intelligence to forecast which members of the new customer group are likely to follow the same path to brand loyalty, it provides a whole new way of thinking about which customers to invest in over the long run – not just what to do next.
Similarly, consider your active customers, who are no longer new and have purchased multiple times from you, illustrating some level of engagement with your brand. You could use predictive intelligence to ask a related question to the above: Based on this customer’s pattern of interactions with my brand, whether transactions or other measurable attributes of behavior, what is the probability that this customer will become one of my best customers? Predictive modeling can also help you from the other direction: Based on their observed pattern of interactions with the brand, which of my active customers are exhibiting a high likelihood of lapsing? These customers may need a little extra encouragement to stay connected with your brand, and predictive intelligence can help you identify who they are so that you can test various tactics to retain them.
Jim Sawyer is Chief Scientist at Elicit. The company's resident savant, Jim is responsible for the artistic application of Elicit’s customer science. From evaluating the state of customer data and analytics systems to developing customized segmentation, Jim leads a team of data scientists to bring customers to life through data. He has over 20 years of experience in analytics, a Stanford B.A.S. in Mathematical and Computational Sciences and a Georgia Tech M.S. in Industrial and Systems Engineering. Elicit's Fortune 500 clients include Southwest Airlines, Fossil, GameStop, Sephora, BevMo!, HomeAway, Best Buy and Pier 1 Imports.