Using AI to help Red Cross Donors Leave a Lasting Legacy

In 2021 Australian Red Cross launched a new bequest program, “Friends of Red Cross Society”. The aim was to encourage more supporters to leave a legacy of support for people experiencing vulnerability.

Understanding that one of the best places to find new bequestors is within your existing supporter base, Red Cross’ data team built a model to help. The model identified potential bequest donors to steward. This added more than 4,000 new bequestors into their pipeline across FY 2021 and 2022.

In June 2023, Red Cross partnered with Dataro to see how our predictive modeling compared to its in-house approach. Red Cross also wanted to know whether Dataro could help uncover additional bequest prospects.

In the first month of testing, the Dataro model accurately identified 100% of the 11 newly confirmed bequestors. The model also helped convert more bequestors to the pipeline, helping to expand the program further.

Key Results

In the first month of testing:

• The Dataro model accurately identified 100% of confirmed bequestors
• 80% of intending and 74% of considerers came from the Dataro model
• 11 bequest gifts confirmed by the Dataro model in one month

How we did it

Red Cross runs a monthly bequest campaign, contacting approximately 42,000 “warm” donors annually. This number fluctuates during critical appeal periods. The campaign uses direct mail as the first contact point, followed by telemarketing to those who don’t convert during the initial outreach. 

Red Cross wanted to know if Dataro’s machine learning model would improve results by identifying additional prospects to target. 

 Specifically, the charity wanted to know whether donors identified by Dataro as good potential prospects led to more bequest confirmations. And if that would add more donors to their pipeline at higher rates than if Red Cross only used its in-house model.

The test compared selections from the Dataro model against the Red Cross model to identify:

• Donors in the Dataro group only (2,500)
• Donors in both groups (500)
• Donors in the ARC group only (2,500) 

Many outcomes that fundraisers are interested in, such as bequest confirmation, are exceedingly rare. For instance, bequest gifts account for less than 0.05% of all gifts. A critical aspect of the success of any machine learning system is that the event being modeled occurs with sufficient frequency to accurately estimate the relationship between the model factors and the outcome. As such, in Dataro’s experience, most in-house attempts to model rare events such as bequests are likely to produce inaccurate predictions. This is a factor even at large organisations with a deep history of bequest gifts.

Dataro has been able to overcome this difficulty by training ‘foundation’ models for rare fundraising events based on a data pool of over 100 organisations. Using special sampling and validation techniques, we are able to provide sufficient positive cases to the model in training to attain very accurate predictions from the model.

Using Dataro’s stewardship model, donors are ranked based on their likelihood to be interested in leaving a bequest. These scores are integrated within the charity’s CRM against the donor record as a donor rank, with 1 being the most likely to confirm a bequest gift. This allows the fundraising team to identify the best prospects to include in their fundraising campaigns.

Analyzing the results

This test revealed that Dataro’s machine learning model was accurate in identifying prospects who were likely to leave a bequest, with 11 confirmed gifts identified by Dataro in the first month. Dataro’s selections also uncovered 80% of intenders and 74% of considerers added to the pipeline, highlighting the effectiveness of Dataro’s approach and also the value of using the models side-by-side.

Red Cross continues testing the Dataro model alongside its in-house model to make donor selections for its monthly conversion campaign.

As of 30 June 2023, Red Cross has added 2,400 new bequestors into the pipeline, 40% up on the previous year. The results show that AI can be highly effective at identifying donors interested in leaving a bequest. The direct mail and telemarketing strategy has also proven an effective way to convert Red Cross donors and help them leave a legacy of support for people experiencing vulnerability.