In our recent Water Cooler Chat, we talked about using sentiment analysis for getting a glimpse into the minds of our constituencies. For instance, is there a way to understand our donor clusters? Do donors to the new computer lab like to give cryptocurrency, for instance? Gift data itself is the most tracked data that we have, since it is audited. This blog post explores more deeply the idea of uncovering donor interest through relationships among pieces of giving data.
Relationship mapping helps identify natural clusters of constituent characteristics, which can then be used to form segmented solicitation efforts. In other words, moving from a “September direct mail” campaign to an “Online donors who give in September to the computer lab” campaign would help us connect much more personally to members of our audience.
In the Water Cooler Chat, I showed the machine learning technique, cluster analysis. Here is the illustration from my presentation.
If you’re not interested in this much (or this complex) data preparation and analysis, consider using NodeXL as a clustering tool. This time you are using NodeXL to map relationships to data points instead of to other people or social media accounts. Let’s take a look.
First, every data exercise starts with data. Let’s say that we have this data on last year’s annual giving program:
Month of last gift date
Last fund given to
Last solicitation method given to
What can we learn by using NodeXL’s relationship mapping exercise? Can we identify clusters of giving behaviors?
First, we would load the data into NodeXL. Below I show a made-up data set loaded into NodeXL, and created my first graph, which didn’t tell me much:
When I summed up the data to count the repeated rows in order to get a clearer picture, I saw that most gifts are to the unrestricted fund, as shown below.
Adding the solicitation method as the edge color shows me that email is a big attractor for unrestricted gifts.
The other funds which I named “Jump Rope Program” and “Building Fund” receive more gifts from direct mail. Let’s take a look at that by excluding the unrestricted gifts.
(Red values are the Jump Rope Program.)
When I recolor the graph to represent the average gift amount, I get higher gifts from direct mail to the Building Fund. See this next graph.
So next year, direct mail will be sent to prospects for the Building Fund, and the first solicitation for unrestricted gifts will be sent via email.
Have we shown sentiment? Yes. We’ve shown that email donors are responding to unrestricted appeals and that specific fund donors like to be solicited in a more personal way, even if that is only a direct mail piece.
Imagine if our 4 pieces of data included posts responded to or liked, key words from those posts, or count of events attended before the gift. What will you find?
Give it a try! And I hope that you will share your results with us.