In the eight months since incorporating Vasker, I’ve done and learned a lot. It’s been an intermittently exhilarating, discouraging, focused, wandering process. Here are some things I’ve learned, done, and thought about.

  1. While ability and affinity are important measures of a prospect, past giving is considered the biggest indicator of future giving. Beta clients in two states brought this up, leading me to import gigs of historical political donation data from the public sources.
  2. There’s no perfect way to associate donor data with voter data. You can match on names, but there are problems, including the Bill/Billy/Will/William issue and the fact that there are many people named Bill Williams in any city. But you can make a best effort and display the information as “possible donations” to be reviewed by human researchers.
  3. Speaking of names, people’s names come in all sorts of forms, and sometimes they don’t use their real names, so you need to be able to recognize “Dr. Bill E. Williams, PhD” as particular voter #12344567 “WILLAM EDWARD WILLIAMS”, while discarding “Mr. Bill is Awesome, Yo”. External databases of common words, nicknames, surnames, and suffixes can help. But watch out — there really are people whose names consist of dictionary words, like “Will Hurt”, “Bill Gates”, “Dick Smith”, “Bob Black”, “Melody Winters”, “Grace Masters”, etc.
  4. Earth is not a perfect sphere, but it’s close enough at most latitudes that simple geometric calculations suffice for distance calculations between points, rather than GIS science, of which I know almost nothing.
  5. Professionals who work in politics (fundraising, managing data, coordinating volunteers, handling communications) are like mercenaries: They go where the fight is. So your contact within a political organization may not be around long. My hunch is that this enhances the importance of professional networks and reputation in the industry. People move around, crisscrossing the country and each other’s paths, over multiple years and election cycles.
  6. Some statewide political organizations tell me “we don’t do prospecting”. They already have a deep list and don’t have or require staff to conduct prospect research. Others tell me they have a great donor list, but a high percentage of them are people over eighty years old, and so they are actively prospecting for younger professionals to activate, and have dedicated staff to support that effort.
  7. My system for measuring affinity or interest scores individuals based on their specific expressions of personal interest. The sophisticated scoring used by the Democrats to rate a voter’s likelihood of voting Democratic takes into account is a predictive model, rather than an individual score.  I ran a Pearson Correlation to see how well the two scores line up, and got a moderate positive correlation.  I am not sure whether that’s a valid comparison or if these two scores are different enough in kind that it’s meaningless to compare them statistically.
  8. I aspired to use git for source control at the beginning of this project, but soon began ignoring it as a time-wasting step.  This hasn’t burned me – yet.
  9. I am convinced that my software is identifying real people who lean politically progressive. For fundraisers (my clients), the question remains, what’s the best way to activate them as contributors? How do you create a first-time political contributor? (It’s a difficult topic to Google, since there’s so much news about campaign finance and political contributions. One essay posits outrage as the best tool to activate new contributors, but does so in a pretty disparaging way.)
  10. I’m awaiting results from a test — 5000 prospect phone numbers sent to a call center to be read a four-sentence fundraising script. Is this the best way to activate them?
  11. This is the biggest software/data project I’ve ever taken on single-handed. I valued and miss collaboration with founders/employees. And, as I’ve been forced to learn new technologies and make software design decisions, I’ve levelled up, which feels good.