At the end of August, news broke about a record-setting drug bust in Australia that caught world-wide attention. Three Canadians were allegedly caught trying to smuggle $23M worth of cocaine (about 200lbs!) into Australia via a cruise ship they had been traveling on. What’s the connection between this amazing story and the world of fundraising analytics, you ask? Well, aside from being excellent source material for a made-for-TV movie, this story is a perfect example of a concept that is central to predictive modeling: the difference between data that Describes versus data that Distinguishes.

picture1

 

For a little background into this intriguing story, the three people charged were a 22 year-old woman, a 28 year-old woman, and a 63 year-old man. While it wasn’t immediately clear whether or not they were traveling as a group, reports indicate they had been cruising around the globe for nearly two months on a $15,000 per person luxury cruise that took them to the United Kingdom, Canada, the US, Colombia, Peru, New Zealand and Australia.

The most interesting part is that the suspects had been identified as “high-risk travelers” through a cooperative effort between US, Canadian, New Zealand, and Australian law enforcement. We don’t know all the details, but we can make some guesses and clearly illustrate the concept of Descriptive versus Distinguishing in the process:

According to US cruise industry market data available at www.cruisemarketwatch.com, individually, our three suspected drug smugglers were anomalies; if they were indeed traveling together, they must have stuck out like sore thumbs to law enforcement agents! (Be sure to check out the great Tableau-driven analysis embedded in the site!) Using this data, we’ve compiled a short list of some of the ways in which our three suspects Distinguished themselves from a typical cruise ship passenger, leading to their categorization as “high-risk,” and ultimately their arrest:

  • Age: Two of the suspects are under 30, but people under 30 make up only 7% of all cruise ship passengers.
  • Marital Status: 78% of passengers are married; it appears all three suspects were single.
  • Employment status: Assuming the female suspects were employed full-time also serves to distinguish them from typical passengers as retirees are much more likely to cruise.
  • The cruise itself: How many people in their 20s do you know that can afford a 2-month, $15,000 cruise around the world?
  • The group: If all three suspects were traveling together, their group composition may also have worked against them to alert police. Two young, single women traveling with a 63 year-old man, all of whom are unrelated to one another is likely quite unusual.

The same techniques are required for successful predictive modeling in fundraising. We don’t want to Describe major gift donors; we want to identify the characteristics which serve to Distinguish them from the rest of our constituent population. Although describing certain groups can be beneficial in order to understand more about their composition and perhaps their motivations, this type of analysis does not help us uncover more people like them. Being able to say that 75% of our major donors are married only describes them, it does not distinguish them from other groups (unless we find that only 40% of our entire constituency is married, in which case the 75% figure among major donors is very useful in distinguishing them).

The difference between Describing the characteristics of a group and using data to Distinguish one group from another is subtle but extremely important. This concept serves as an important foundation for exploring the possibilities of predictive modeling with many of our clients, particularly those new to analytics. Maybe if our three suspects had known a little more about the power of predictive modeling, they wouldn’t have found themselves in their current situation!

To learn more about analytics at BWF Insight click here!

Share This