A sophisticated analysis by Stanford data scientists supports a common sense approach: help fundraisers craft successful appeals and thank givers promptly, and they will return to give more.
BY TOM ABATE – Stanford computer scientists have shown how crowdfunding websites can use data science to boost cash value of donations. Their research confirms, among other findings, the importance of a timely thank you.
It’s common courtesy to say thanks when receiving a gift, and common sense to think that givers might be generous again if they felt good about their prior gift.
Now Stanford computer scientists have shown how crowdfunding websites like Kickstarter and DonorsChoose.org can use data science to apply these folk wisdoms systematically, and on a large scale, to greatly boost the cash value of donations.
In a paper for the International World Wide Web Conference, Stanford computer science Assistant Professor Jure Leskovec and graduate student Tim Althoff analyzed millions of online donations collected over a 14-year period.
Their findings suggest that increasing the number of repeat givers by just 10 percent could yield over 60 percent more dollars donated over time.
“Most people give once and never return,” Althoff said. “When you increase the number of repeat givers, you get a multiplier effect. The people who become repeat givers tend to give more with successive donations, and they also recruit others to give.”
Their analysis of anonymous data provided by DonorsChoose.org, a crowdfunding site that helps educators raise money for class projects, also reveals some of the ways to increase donor retention. These include such suggestions as posting smaller requests, and how to time “thank yous” and project reports.
The researchers say the findings from the DonorsChoose.org case study will help other crowdfunding sites, as well traditional fundraising organizations.
Leskovec said, “What excites me as a computer scientist is the unprecedented opportunity to learn about human behavior by observing things on a large scale. In this case we had excellent data that enabled us to discover new tools and techniques for the fundraising community.”
The study data were provided by DonorsChoose.org, a website that enables teachers to request donations for class projects or activities that they can’t obtain otherwise. Between 2000 and 2014, teachers have posted 638,000 projects and raised a total of $282 million.
Over this 14-year period, the organization received 3.9 million donations from 1.5 million donors. But the Stanford analysis focused on 470,000 people who donated between 2009 and 2014, when DonorsChoose.org started operating nationwide.
Of these donors, roughly three-quarters gave once and never gave again. Just 26 percent gave at least a second time. And only 1 percent of these studied donors gave five times. Traditional nonprofit organizations and charities report similar donor retention numbers.
The computer scientists analyzed the data in many different ways to arrive at some general rules to up those percentages. For instance, was the donor giving to a project in their own neighborhood? Did the donor’s first project succeed in meeting its fundraising goal? (If not, the donations were returned.) And, did the teacher send out thank-you notes and/or project reports, and if so, when?
These analyses sought to discover what sort of factors favored repeat giving. In broad strokes, smaller projects were more likely to succeed, and donors were more likely to give again if their first gift contributed to success – suggesting that teachers start with modest projects.
When the gift was acknowledged was another key consideration. Almost all teachers eventually sent a thank you after their project was funded, but retention was higher when this acknowledgment came during the first hours after funding was achieved.
These and other findings helped the Stanford researchers build a model that included 21 different variables. They used this model to help predict which donors would be most likely to return – for instance, those whose first project succeeded and who got prompt acknowledgment of that success.
Building this predictive model was a key contribution of the research because it is designed to help fundraisers focus on cultivating people who are more likely to become repeat donors.
“All fundraisers have to focus their efforts,” Althoff said. “What we are doing with this study is figuring out what data you should collect and how to start using that data to identify your most likely return givers. This is where your scarce campaign dollars are best spent.”
In this study, DonorsChoose.org not only furnished the researchers with data but also sought to understand how to use the findings to improve their platform in several ways. For instance, the organization encourages teachers to expedite the “thank-you note” by suggesting they pre-write such a note when they originally post their project.
As for estimating the potential to increase fundraising based on boosting donor retention, the computer scientists used their model to project what would be the result of pushing donor retention from 26 percent to 36 percent. They found that a 60 percent increase in total dollars donated was likely as repeat givers tend to up their gifts and also recruit their friends.
Tom Abate, associate director of communications, Stanford Engineering: (650) 736-2245, firstname.lastname@example.org
Dan Stober, Stanford News Service: (650) 721-6965, email@example.com