Dear {{first_name|friend}},
I am tired of “think” pieces about how Artificial Intelligence is going to affect jobs and the economy. Doom and gloom, or self-motivated hope-mongering, which flavor are you most bored with?
I’m sick of both! Instead, I want to share something from research Chorus has completed. It got me thinking about an unexpected change from AI tools gaining adoption.
Chorus partners with nonprofits and the businesses that work with them to understand their audiences at a new level. Unlike tools that look just at past giving behavior, Chorus analyzes content (topics, themes, emotions, framings) and context (news, timing, channel, social media environment) at the level of each individual person. Our goal is to help our partners raise more money and achieve more engagement through deeply personalized marketing.
We’ve conducted a few large-scale studies with businesses like Civic Shout, an ad platform with one of the strongest advocacy email programs in the progressive space, and a Nonprofit with an annual budget over one billion dollars that works in the human services area.
What we discovered over thousands of email sends to millions of people: supporters have distinct topic and content preferences. For the human services Nonprofit, email routing – meaning getting the right specific topic to supporters based on their preferences – can improve the email donation rate by 2.25x. In Civic Shout’s case, content routing for email petitions in a randomized controlled trial (RCT) lifted the signing rate by about 3 percentage points (around 10% more signatures) and clicks by about 18%.
No duh, right? For as massive as these improvements are, they are also kind of commonsense. Thinking about my own giving, I’m more likely to donate to a conservation fundraising ask than a clean energy one from an environmental nonprofit. But the reality is most nonprofits send one email or SMS to everyone, and the reason isn't intellectual — it's that doing the work is a complete pain in the ass! Setting up twelve variations of a Giving Tuesday email, generating the copy, cutting the list, syncing the sends, measuring what worked, and feeding it back into next week's targeting is more than most teams can take on at the standard ROI you would expect.
But that’s where new technology comes in. Vector embedding models can now turn a piece of writing into a numeric representation that captures not just its words but its meaning, and the resolution has gotten fine enough to tell closely related topics apart. (This is the same fundamental technology underpinning large language models.) Combined with data science and machine learning, we can actually understand every audience member as an individual, and we can realistically route them a message they are more likely to respond to.
We can’t predict what this means for nonprofit digital and fundraising teams with certainty. I’ll still hazard one theory: it means writing a lot more content. The economics of writing two or three fundraising emails monthly, to say nothing of tentpole times of the year like Giving Tuesday or end of year, are different when those emails can massively outperform the baseline. Rather than cutting back on digital teams as new technology and tools are woven into nonprofit operations, perhaps we’ll need bigger teams to write more copy. Our supporters can receive the same or even less content, but far better targeted to their own stories and interests.
I suppose this is hopeful, that AI won’t mean alienation from nonprofits and charities and mission-driven organizations. Instead, our research suggests it can be a component of building deeper relationships and providing meaningful news and information, opportunities to give, and calls to volunteer. But we’re going to need a lot more writers.
Thanks,
Sam
P.S. Want to learn more about Chorus or read the results of our research in detail? Reply to this email or book a walkthrough.