When I’m about to begin work on a new predictive model, I get a little scared. When I agree to speak to an audience with more brains and experience than I have (i.e., most any audience), I get a little scared. Heck, when I’m about to click the ‘Publish’ button on a new blog post, I’m a little scared.
But here’s what I’ve learned from working on unproven models, meeting and talking to smart people, and writing about stuff: When we feel a little bit scared, it means we’re probably onto something. It’s a signal to press on, not duck for cover.
Think about advanced data work. Here’s the thing: It’s totally optional. In fundraising or marketing or business, you can bull your way through without it. Can a fundraising program be successful without predictive modeling? Yes, it can.
I’ve heard it said that success in Annual Fund is based on making many tiny, incremental improvements over time. A gain of half a percent here, half a percent there will add up to substantial progress. These are little tweaks to familiar variables: the form and content of an appeal letter, the shape and size of an envelope, the choice of which Phonathon attempt to leave a message on, et cetera.
We fundraisers are, I think, cautious and conservative. We are very receptive to the idea that continuous progress is possible without having to learn anything new. I’m cautious and conservative myself, so I get it.
I just don’t find it very interesting.
We can probably do A/B testing forever, and we should. But at some point there will be a limit on returns. When steady application of the tired/tried-and-true fails to result in a year-over-year gain, we need to stop blaming external factors such as the economy or major world disasters (significant though they may be) and get serious about how we focus our efforts.
Judging from the questions I see on some of the listservs, fundraisers are being challenged to stretch. Some are responding by exploring boldly, most are just rearranging the known elements. Hardly ever does someone suggest doing serious work with data.
Working with predictive models is as iterative a process as any of the traditional stuff, once you’ve gotten started in that direction. Your models and predictions will get more focused year after year with incremental improvements in data collecting and analysis.
But getting started is not an increment. Getting started is something new, and starting something new is a little scary. Do it.