Dataro and Greenpeace Australia Pacific reveal how they are using machine learning to improve donor engagement and fundraising performance.
Corporate Australia is no stranger to artificial intelligence. Recent research by Deloitte suggests 59% of organisations are already using some form of ‘AI-as-a-service’ technology and 79% believe AI will be ‘very’ or ‘critically’ important to their business within two years. Now the charitable sector is catching on, with early adopters beginning to realise the benefits of the game-changing technology.
Greenpeace Australia Pacific is one of the thought leaders investigating how machine learning can be used within its fundraising and other programs. The environmental organisation has been working with Sydney tech startup Dataro to test and deploy machine learning predictive modelling for a range of fundraising campaigns, including regular giver churn, direct mail appeals and lead scoring.
According to Dataro CEO and co-founder, Dr Tim Paris, early results show how machine learning could shake up fundraising in the same way it has revolutionised marketing in the commercial sector.
“Early adopters like Greenpeace are helping to define a new best practice in fundraising,” Dr Paris says. “Already we’ve found that machine learning algorithms have a much greater ability to predict donor behaviours than traditional methods. We’ve had results including 2.5 times more regular giving reactivations in a call campaign, and reductions in mail volumes by 50%.”
Machine learning is the label given to a field of computer science where computers are taught to analyse large amounts of data, identify complex and nuanced patterns, and predict future outcomes without explicitly programmed instructions. The systems are designed to ‘learn’ and improve in accuracy over time when exposed to new data.
Dataro and Greenpeace first teamed up in 2018 to test how well machine learning could be used to predict regular giving churn. “We ran a controlled experiment and found that not only were we able to accurately predict which regular donors were the most likely to churn, but that Greenpeace could actually improve regular giving retention by 15% over six months through engagement calling,” Dr Paris says.
Like all AI technologies, machine learning relies on access to useful data. For most charities, the data stored on their CRMs is normally sufficient.
Dr Paris explains that for the best results, the full history of an organisation’s fundraising is used, taking into account potentially hundreds of factors ranging from transactional and communication histories through to card declines, email activity, age, time of year and more. This data is analysed using a range of algorithms to identify how important each factor is and how they can be combined mathematically to produce models that predict the likelihood of a particular outcome. To verify the quality of the predictions, each model is tested against withheld data to see how well it performs, giving a real indication of likely future performance. The best model is then deployed to generate predictions to use in the campaign.
Dr Paris says the result is that each donor receives a score indicating their probability of participating in a campaign, allowing charities to refine their campaign lists and choose better list sizes by removing donors with lower scores and adding donors with higher scores.
Dataro and Greenpeace joined forces again for the organisation’s tax appeal, this time to test how effectively machine learning could be used to predict appeal gifts. As with many charities, direct mail appeals are one of the planks in Greenpeace’s fundraising strategy. Dataro generated predictive scores for every donor in the database, and then compared the scores against the actual campaign results.
The results demonstrated the enormous potential of the technology to predict donor behaviour. Donors with the highest scores had the highest direct mail response rates, the best ROI, and contributed the most revenue. Further analysis showed these donors also contributed the most out-of-campaign revenue during the campaign period, which may be attributable at least in part to the mail campaign. Based on such results, Dataro and Greenpeace are rolling out new models for acquisition lead scoring, regular giving fundraising and future direct mail appeals.
“Machine learning is here to stay, and these results again show how new technologies can deliver a better outcome than traditional segmentation methods like recency, frequency and monetary value analysis,” says Dr Paris. “These trends are typical of what we see using a machine learning approach. We hope that by using this technology we can empower our partners to create smarter campaign lists and make better fundraising decisions, ultimately raising more funds for their causes and improving their relationships with their supporters.”
For more information visit https://dataro.io/