A Data Science Case Study: Finding and Keeping Donors

In my recent article series on building a high-performing data science team a theme has emerged from the feedback: give an example of how data science and data scientists can impact a business. This article is a case study of work Bluejacket has done with a fine arts non-profit. The non-profit had an executive leadership team, fundraisers, marketers, and data analysts on staff. Their data was relatively well organized and centralized in a customer relationship management (CRM) system. This article will address some of the keys to a data science project:

• How are the problems defined? Do all three--get, keep, grow--need to be addressed?
• What are some of the classic data science problems?
• What is "relevant" data?
• How do you define the scope and refresh rate of a data science project?
• What is the output from the data science team?
• How is that output leveraged by the business in an actionable way? 

The non-profit had a dedicated and, in years past, successful staff of fundraisers. But contributions were down and some existing donors were cutting back or not contributing at all. The non-profit had two problems:

1. Existing previous donors are decreasing or stopping their contributions
2. The overall donor base is stagnating; new donors are necessary for an infusion

The two problems may seem obvious to you and they may not. The first problem is, in many industries, known as a "churn" issue. Churn is when existing clients or customers leave or stop buying a product or using a service. Churn is a classic problem for a data science solution. The problem of finding new contributors is acquisition, targeting or "get", and is not as often considered as the churn problem, though they're often two sides of the same coin. Many organizations don't consider acquisition as the churn complement when they should.

There's another addressable issue: increasing existing steady contributor donations. That's the 3rd leg of the get, keep, grow stool to which data science can contribute. Most organizations have three core issues: adding new customers, keeping existing customers, and growing existing customer relationships. Historical data and data science are keys to developing successful action to improve all three.

In our example the 3rd issue of increasing existing contributor donations was proposed during the discovery phase. In this case, though, the idea was turned down because of the urgency to focus on keeping existing donors and returning them to former contribution levels and the opportunity to find untapped donors.

We agreed that the data science output would be two prioritized lists: the first would be a list of existing donors whose contributions had been decreasing over the last three years. The second list was potential donors in the same geography who had never contributed but had attended functions in the last three years or were potential attendees. These lists would be used by fundraisers and marketing to prioritize visits and marketing outreach as part of the donation campaign for the coming year. Success would be determined by comparing existing donor contributions from the previous three years and comparing new donation rates and the number of new donors to the past three years.

The non-profit had more than 20 years of data on past contributors, which included demographic and financial detail in the CRM. In ferreting out details in partnership with the leadership team we determined that related, relevant data associated with the contributors included a history of contacts such as personal visits by fundraisers, event attendance, emails, and hard copy mailers. The contact data was spread across the organization but primarily found in departmental databases and spreadsheets. A single, unified view of contributors including their demographic, financial, and contact history was not available but became a goal of the project.

Developing the most-complete contributor "profile" is critical to the predictive modeling process.

• What if certain contributors did or didn't increase or at least steadily continue when they were visited by a fundraiser?
• Or if they did or didn't receive email or a flyer?
• How does the combination of these touchpoints improve the likelihood the contributor would increase their contributions?
• Have contributors changed jobs or positions recently that could impact their ability to contribute?
• Does contributor recent event attendance impact the probability of their continued contribution?
• Has there been a change in the personal lives of the donor, such as an illness, child going off to college, that might alter the contribution that may have been noted in the CRM record?

The final output, the two lists of donors and potential donors, were delivered and integrated into the call and visit patterns of fundraisers and the development of hard copy and digital marketing materials. A set of champion/challenger (AKA A/B tests) were also developed to measure the success of the new campaign versus previous campaigns. As of this writing, while event attendance has remained statistically stagnant the donation campaign has reduced overall marketing costs by more than 20% and donations are up approximately 11%.

Future projects to enhance fundraising that have been proposed include a network analysis of donors and potential donors and a natural language analysis of marketing materials to determine the impact of language on donors' probability of contributing. Network analysis is an exciting and not-well-known aspect of data science used by organizations such as LinkedIn, Facebook, and Google to determine the connections and strength of those connections between people. For example, people are often influenced by who they know to get information or feedback before making a purchase. Natural language processing can determine the impact individual words, phrases, or statements have on the likelihood of someone taking action. Determining the language, tone, and theme of your target audience is one way to positively impact the likelihood of reaching them and having them take the action you want.

Much of the work between the development of donor profiles and the selection of the best model, the explanation of the factors that impacted donations, and the next agreed upon steps between us and our client before the production of the final lists have been neglected here. Those steps are critical to the success of the project, some of which are: milestone discussions of progress, confirmation that the analysis is moving toward the agreed upon output, explanation of findings that offer an opportunity to confirm that direction, and access to our stakeholders as partners in their business. 

If you've enjoyed this example of the impact of data science and the steps for getting there drop me a line at sam.johnson@bluejacketsol.com. If you're ready to assess your data science practice or your organization's readiness and/or need for data science, don't hesitate to reach out. If you'd like me to address questions you have about data science in other articles, reach out. Bluejacket has more than 10 years experience helping organizations determine their needs and readiness for the successful integration of data science to create competitive advantage. We use data science to build better data science.