Joining the fight against prostate cancer with Prostate Cancer UK

Dataro recently joined forces with the team at Prostate Cancer UK to test how AI predictive modelling would perform against traditional list selection methods in an Autumn Appeal. The results showed that Dataro’s selections increased revenue by over £11,000 and attracted more than 440 new gifts that otherwise would have been missed, laying a solid foundation for future appeals.

Key results

• Over 440 extra gifts that would have been missed
• Over £11,000 additional raised
• Significant increase in response rate

How we did it

Prostate Cancer United Kingdom is innovating in the way it uses data to select which supporters to contact with which campaigns. Traditionally, charities have used simple models or manual methods such as RFM (Recency, Frequency, Monetary value) analysis to create segments that group supporters together based on recent giving history.

Dataro’s software uses machine learning to analyse patterns of donor behaviour across all fundraising campaigns and understand each donor’s unique giving journey. Each donor is then given a ‘propensity score’ reflecting their probability of giving in an upcoming appeal. This process was conducted using Dataro’s integration with Blackbaud Raiser’s Edge CRM.

In this experiment, we compared the results from both approaches using the same methodology as with organisations such as Parkinson’s UK and Greenpeace:

1. We generated propensity scores for every donor reflecting their propensity to give to the direct mail appeal. Using these scores, we generated a recommended campaign list.
2. Prostate Cancer UK compared the Dataro list against its own proposed campaign list to identify three categories of donor:
   1. Donors appearing in both lists were labelled blue;
   2. Donors Prostate Cancer UK selected but which Dataro would removed were labelled red; and
   3. Donors Dataro selected but which Prostate Cancer UK had not chosen were labelled green.
3. The appeal was sent to all donors from both lists, allowing us to compare responses from the Dataro selections vs the Prostate Cancer UK selections.

Analysing the Results

The results were clear. Dataro’s machine learning list performed better than the classic list across all metrics. Critically, the results showed:

• Organisations not using machine learning are missing out on a significant number of donations in their appeals
• Organisations typically include far too many donors, resulting in wasted mail and increased cost

Net Revenue

A key test of performance is net revenue. In this case, net revenue takes into account the costs of the campaign and the total amount raised. We could see that the net return from the blue segment was the highest, but that the extra donors selected by Dataro significantly outperformed the extra donors selected using the classic model. This resulted in an overall 3.6% increase in net returns.

Response Rate

Response rates are useful to track the effectiveness of list selections. In this analysis, we can see how the extra donors selected by Dataro responded far better than the extra donors selected with RFM, with a response rate of over 8% compared to less than 3%. The donors appearing in both lists also performed strongly, with a response rate over 10%.

ROI

A final key indicator for the campaign is the ROI, which in this case takes into account the total mail cost and the return from the Dataro list vs the classic segmentation. The results showed an ROI of over 10 from the Dataro campaign compared to 8.7 from the normal selection, an increase of 14.8%.

These results again show that better campaign selections driven by machine learning predictive modelling result in more gifts and increased returns in direct mail appeals, paving the way towards higher performing appeals in the future.