Expanding a regular giving program is not always about acquiring completely new donors. In fact, sometimes the best leads might be right in front of you in your lapsed donor pool.

Dataro has achieved strong results using our machine learning approach to predict which regular givers are the most likely to reactivate. In this article, we take a look at some of the important model features we take into account when generating reactivation predictions and some of our key reactivation findings.

Predicting reactivations is a complex task. By definition there is no recent regular giving history, which means classic segmentation approaches like RFM analysis will not work. The solution charities use at present is to deploy a few simple rules, such as calling all lapsed donors six months after they churn. But this approach fails to make the most of the rich data that charities have about their lapsed donors, such as age, giving and communications history. In fact, in our experience time is almost never the sole dominant factor. The advantage of a machine learning approach is that it takes all of that data and more into account to generate predictions for each individual, resulting in more targeted and successful reactivation calls.

Giving History

Perhaps unsurprisingly, an individual’s giving history is an important indicator of their likelihood to reactivate. The most loyal donors are also often the most likely to reactivate at a later date.

In these charts, we can see how the number of previous gifts, number of previous upgrades, and total donated are all important indicators of a donor’s propensity to reactivate. In each case, donors with more previous gifts and upgrades, and higher total donated, were more likely to reactivate. Interestingly, however, prior average giving amount tended to have a smaller effect.

Number of previous gifts

Number of previous upgrades

Total donated

Average gift amount

Time since churn

Many telemarketing reactivation programs target donors based on a certain period of time since that donor churned. While in some cases more recent churners may also be more likely to reactivate, this is not always the case. In this graph, for example, we see that people who churned between 1-4 years ago were the most likely to reactivate, whereas people who churned less than 1 year ago were about as likely to reactivate as long-lapsed donors.

Tenure

Generally, mid to long term donors seem to demonstrate a greater willingness to reactivate a regular gift. Again, this is not the same for all organisations, but in this example we can see that donors who churned within 1 year were substantially less likely to reactivate their regular gift than donors who had remained active for 1 – 10 years. 

Conclusion

There are many factors relevant to predicting regular giving reactivations. Understanding how they interrelate in your program specifically is a complex task. It is clear, however, that simplistic methods such as applying a ‘time since churn’ rule are inadequate. In our experience, far healthier reactivation results can be achieved through the use of data science solutions that predict which donors are likely to respond at any given point in time.