Glimpsing Obama Campaign Fundraising Segmentation and Micro-Targeting

ProPublica has tried to reverse-engineer an Obama fundraising email, different versions of which went to a husband and wife:

Sinker and his wife weren’t the only ones to receive similar but subtly different emails from the Obama campaign. Responding to a call on Twitter from Sinker (and another from us), 190 people from 31 states and Washington, D.C., sent us the messages they received.

A look at those emails shows the campaign sent out at least six distinct versions of the fundraising appeal.

The reasons for the differences remain unclear. (The campaign hasn’t responded to our requests for comment.) It could the campaign testing which phrasing gets the best response. The messages may also be tailored to individual voters based on the campaign’s extensive database of personal information.

Either way, it’s a glimpse into the detailed data work that rarely gets attention but is increasingly central to campaigns.

The cool part? They’ve posted the variants online and tracked the variables among them as best they could, so that we can try to deduce the underlying logic. A fun party game for digital politics nerds, most def.

Welcome to the age of database-enabled communications! I say this as a professional online communicator who had three different variants of a fundraising email sent out on the same day last week — one targeted people who had donated to the organization in the past, while the other two tested a $10 ask versus a $20 ask for people who hadn’t donated before (the $10 variant raised far more than the $20 variant, btw, while also adding three times as many donors to our list).

This kind of variation is extremely common in the advocacy and political worlds these days. For instance, a campaign might send out a cluster of different appeals all at once, with different people getting different content based on which appeals or actions they’d responed to in the past. Or, the campaign test how different segments of their list respond to different messaging or different amounts. A common trick: to tie the ask to someone’s previous donation amount, always asking for a little more than the last time.

As you might guess, the potential variables available for testing are nearly infinite. How long is the email? What about associated imagery? How far down in the text is the ask? What about different subject lines? A complex endeavor — and one that’s really hard to track from outside the campaign, since there are only so many email accounts you can manage. Some of the first words on Epolitics.com? The real work of electronic politics may be taking place behind the scenes — something that’s at least as true now as it was when I wrote that in 2006.

– cpd