This article was first published as “Marianne’s Mining Minute” in the APRA Upstate New York chapter’s newsletter, 2010.
We have focused for the past 10 years or so on data mining for major gifts prospects. However, with the advent of measuring social networking, we are starting to see the value of using modeling and mining tools to help annual giving, membership, and events programs. We need to find ways to measure people’s participation with us, even if they participate only by reading our newsletter.
Some measurements are easy, if one has the data: number of events attended (in my own studies, number of events refused was an indicator of major giving, since one indicated a close relationship to my organization by regretting an invitation), most recent event attended, number of contacts received. These measures go along with number of historical addresses tracked and depth of family information tracked: they work if the organization diligently inputs the data.
Some measurements need to be thought out. For instance, distance between the first newsletter sent and the first gift; response to a self mailer (I hate those impersonal things); response to a podcast (how does an organization measure who watched it?); mention of the organization as one’s top philanthropy. Some of these can be known by surveying, but surveying itself raises data entry problems. Some require careful data manipulation.
Indexing the date of the first gift and its amount gives opportunity to compare it to the first event attended, to the first volunteer spoken with (if tracked), and to the distance between the first gift and the fir volunteership (and what if the voluneership came first?). Indexing event dates can help one determine if a prospect likes to go only to, say, opening night, or to events designed for high-end donors only.
Annual giving can rely on RFM analysis to assist with measuring a prospect’s increasing giving, only if the program treats everyone the same way, including moving one up the giving ladder based on triggers. Also, a careful study of how different populations respond to different fund-raising techniques (do engineers and Midwesterners prefer to be solicited by e-mail?) is useful, and their response to the follow up technique is even more useful. For instance, if I am acquired as a donor by direct mail after buying a duck at a duck race, then am I best solicited next by a phonathon, by a personalized letter, or by another event invitation?
What are the stages and their progressions? And how does an organization cipher the different audiences? Like Harry Potter’s constantly moving stairways, an annual giving program would get completely lost to try to segment its population down to pools of, say, 5% of the population in each pool. That’s 20 segments. It’s difficult to write that much copy.
The question of measuring participation falls flat on its face if the data is not there, even there in a way that it can be connected to the central database. For instance, I was able to talk to a museum which measured members’ activities while they swiped their membership cards. I could find out when they arrived, what day they liked to come, whether they bought in the gift shop, and their preferred café lunch hour. The catch is that the museum’s best prospects were giving at the patron level, and therefore walked through the concierge desk to be admitted: no swiping. We could not get data on the very people we cared about the best.
Remember, however, that data gives itself away in many ways. My favorite data example is the Smithsonian. They were trying to measure the most favorite exhibits. It took a while, but they finally figured out that counting the tile replacements in front of exhibits worked. I call that getting there by proxy. What data do you have that can be a shadow of an indicator?