Most data mining and donor modeling projects include one dependent variable compared to several independent variables. Let’s review these concepts briefly before we get into the meat of this article.
The dependent variable represents the question that needs to be answered. Think of it as filling in this sentence: “[What you need to know] depends upon….” The “depends upon” are the independent variables, which we explain next. Examples include a donor’s next gift amount, acceptance or rejection to a request, or relative level of engagement (however you define engagement).
Dependent variables can be categories (usually two, such as “yes” or “no”) or continuous values (such as gift size).
The independent variables are those that demonstrate a relationship with the dependent variable. They answer the remainder of the sentence above. For instance, if the dependent variable is the donor’s next gift, then the sentence would read, “The donor’s next gift is dependent upon [a set of independent variables].”
Very few events or people can be described by one characteristic, and so independent variables tend to number more than one. Some of the models I’ve used have over a dozen.
A sample of independent variables interplaying with the dependent variable looks like this:
“The donor’s next gift is dependent upon the donor’s last gift plus the donor’s age plus whether the donor lives in the same geography as our organization.” In programming terms, the result is actually an equation, since the independent variables are also a assigned a weight to describe their relationship to the dependent variable, and our example would look like this after the equation is determined:
Next gift = (last gift * 1.05) + (age * 507) +(if donor’s geography = our geography then .453 else 0)
Dependent Giving Variables - In data mining, we frequently choose the dependent variable that is easiest to acquire from our primary databases. Most often, we start with the prospect’s lifetime giving. However, we want to be careful when thinking about which giving variable is measuring what we want to see. Here is a smattering of giving variables and their uses.
Lifetime Giving measures only lifetime customer value. In addition, it is often directly related to the length of time a prospect has been involved with your organization, especially for consistent givers. So, lifetime giving values tend to favor older donors.
Giving per Year in System reduces the effect of length of time by dividing lifetime giving by number of years in your system. For instance, a prospect became a member of your organization in 1999 and her giving is $10,000. So, her giving is an average of $1,000 per year for her lifetime ($10,000/10 years). You are still measuring lifetime customer value, but removing the effect of age/length of involvement.
Highest Gift/Pledge measures a donor’s single highest gift or pledge. Note that major gifts tend to be pledges, so it is best to measure pledges for this variable. This is my favorite variable for looking for major gifts prospects.
Velocity or Consistency measures a donor’s recent behavior, whether she is escalating her giving or whether she is giving consistently. I use these variables for annual giving studies. Velocity is calculated as (last full year’s giving/average(previous 3 years giving)).
Distances help measure the amount of required cultivation for different level gifts. Measure the distance from first to first major gift (major gifts), from entry date to first gift (annual giving), or from lowest gift to highest gift (major gifts or high-end annual giving) –hoping that the lowest gift came first.
Which are you interested in? Let us know. Tweet us - @Staupell