Predictive modeling, data mining….too many terms, too little time.
As fundraisers, many of us would readily admit that “statistical stuff” just isn’t our thing. In fact, I can’t even count the number of times I have heard, “I got into fundraising because I can’t do math.” So, where do we find ourselves today? Immersed in a world of techno-speak and number-crunching, a veritable sea of data we thought we were avoiding by choosing this profession.
How do we make sense of it all? Perhaps more importantly, how do we put it to work for us without feeling like we need to get a PhD in statistics? Actually, it isn’t all that mystical and, when you consider that just about every aspect of our everyday lives as consumers is somehow affected by both data mining and predictive modeling, it makes it a little easier to “get”. Consider your last trip to the grocery store. Do you have a little fob on your keychain that was scanned when you checked out? If you, it’s likely you’ve been “modeled”. And you thought you were just getting a free jar of spaghetti sauce!
Technically, data mining is the process of finding patterns among potential fields or variables in a database. Common denominators, if you will, describing a particular type of behavior. For example, if you were to look at a variety of potential characteristics of your best donors, you may find that they have several things in common. Perhaps they played sports, or maybe they attended reunions. These may be helpful things to know as you are getting a feel for what your donors “look” like. Data mining can be done pretty easily to rate affinity. You can simply assign a point system, weighted or unweighted whenever you find the “presence” of a particular characteristic. Adding up the points can help to prioritize prospects, either alone or in conjunction with other screening tools.
Predictive modeling is, on the other hand, just that—predictive. It, too, looks at various potential characteristics within a given constituency. But, rather than simply describing a past behavior, it is designed to predict future behavior. A more complex process than data mining, predictive modeling requires a greater degree of data gathering and validity. All sorts of issues need to be addressed when undertaking predictive modeling. It’s not quite a simple as merely observing the presence of a characteristic. The “analysis” part of the process helps determine when the variable is actually predictive, not just descriptive.
So, when do you use each and which can you do very easily on your own? Stay tuned for Part 2 of this series for examples of ways you can undertake some simple data mining at your organization.
Get nonprofit articles, best practice advice, fundraising ideas and invaluable industry reports and webinars delivered for free!