Business Intelligence Network - The Vision for BI and Beyond

If you were to ask a business person to state growth objectives for revenue, profit and number of customers, they likely would say they want it all and more. In reality, those goals often clash. More customers do indeed equate to more revenue in most cases, but this rarely leads to maximum profit. The reason centers on the cost to serve a customer, which for low revenue-generating customers, often exceeds the margins they produce.

Smart businesses consider the value of a customer or group of customers when determining where to focus their relationship building efforts. An examination of the profitability of a customer often reveals that overall profit is optimized on only 50-70% of the current customer base. Armed with this knowledge, the business can concentrate on understanding the characteristics that differentiate the profitable from the unprofitable customer. They can devote resources to nurturing the profitable relationships, and market to prospects that look like those same valuable customers. We refer to this as customer value management.

One way to facilitate customer value management is through the use of predictive analytics. Models built using this methodology will not only identify the best customers and prospects, but will also suggest what actions to take through an understanding of what variables drive the model.

Consider the example of a for-profit educational institution seeking to maximize program participation, at optimum profit levels. They receive hundreds of thousands of leads annually through television and Internet advertising, as well as field visits to high schools across the United States. Managing the leads and guiding potential students through the financial aid process is costly. The ratio of enrollments-to-leads and starts-to-leads is low. Even after a student starts, withdrawal rates are a significant factor in their profit equation. Historically, the mind-set was “maximize leads to maximize enrollment.” No system was in place to gauge the quality of one lead versus another. A predictive analytics approach changed all that.

Three years of data detailing leads, enrollments and starts, and lead source information (media vs. field) were combined with socio-demographic variables at the census block group (CBG) level to build models predicting both enrollments and starts from leads. The models produce probability scores for likelihood to enroll and likelihood to start for each lead. Scores are ranked from highest probability to lowest and placed in deciles, groupings of 10% of the leads.

Figure 1 demonstrates the effectiveness of the model in identifying high quality versus low quality leads. For enrollments, over 78% come from the top four deciles, or 40% of leads. The success rate is even better for starts, as nearly 92% come from the top four deciles. This has major implications for choosing which leads to pursue first, and which leads not to pursue at all. It is no longer a “one size fits all” proposition.

Because the institution knows the costs associated with processing a lead through various life cycle stages to enrollment and start, they can analyze “what if?” scenarios to find the point at which profit is maximized with little impact on revenue.

Figure 2 shows the relationship between revenue and profit when pursuing leads at different depths of the model. Profit optimizes at the 7th decile, or again, the top 40% of leads. If all leads are pursued, revenue increases by almost 14%, but over 40% of the profit is sacrificed! Conversely, pursuing anything less than the top 40% yields both lower profits and lower revenues.

While the profit story is compelling, most businesses still struggle to grasp the idea of intentionally forgoing significant amounts of revenue. That is where the quality versus quantity aspect of customer value management matters.

Figure 3 illustrates the comparative start rates from leads for deciles 7 through 10 versus deciles 1 through 6. Historically, all leads were pursued equally. Here, we see that leads in deciles 7 through 10 are nearly 9 times as likely to start as those in deciles 1 through 6. Reprioritizing the order in which leads are worked to more heavily pursue high quality leads will quickly replace the revenue loss from the prior example. In fact, just improving the start rate from 4.8% to 5.4% through focused effort will replace all the revenue, while improving profit substantially.

The models that support this customer value management effort produce more than numbers. The top drivers of lead probability at both aggregate and individual lead levels deliver great understanding about how to improve lead conversion. For example, the results indicate:

• Leads from Internet sources are much higher in quality and cost significantly less to generate.
  
  
• Census block groups with higher income levels and comfort with debt produce more and higher quality leads.

With this information, the institution can adjust its marketing spend and targeting and direct its field representatives into high schools in more productive geographic areas.

In this era of such massive computing power and data availability, smart companies choose a quality over quantity approach when practicing customer value management. It is overwhelmingly more profitable.