Data science for fundraisers
- 2. Turn Your Data Into Dollars AKA Data Science for Fundraisers Digi.Raise - June 2019
- 3. ● Alana Richardson UNICEF Australia Digital Marketing Manager ● James Orton Applying data science to the Fundraising Sector through Distil Data Who are we?
- 4. ● A bit history: What is data science? ● A fun game: Age prediction ● Real examples: Natural language processing and predictive targeting ● Practical tips: How to make data science work for my organisation What to expect?
- 5. It evolved! Data Mining Data Analysis Predictive Modelling Statistical Learning Data Science Machine Learning Artificial Intelligence Where did data science come from? Many key data science concepts are built on techniques that have existed for some time. In fact, the first predictive model was built around 1805, it aimed to predict the height of children with tall parents. The inception of data science is hotly debated. I reckon it was about 10 years ago, with a landmark HBR article in 2012. So, it Is a relatively young, rapidly evolving field with a long history!
- 6. ● Breakthroughs in methods and improved accuracy in an uncertain world ● Increasing power of open-source tools like R and Python reduces barriers to entry ● Availability of data = Potential return on investment is now significant for Fundraising Organisations Why is it important to fundraisers now?
- 7. Who doesn’t mind telling us their age? Don’t be shy https://distildata.io/how-we-can-predict-age-from-a-first-name/ A game!
- 8. ● Inspired by similar models in the US and UK. ● Collates publicly available births and deaths data for Australia. ● Calculates likely age distribution of people alive by a given name. Age Prediction: How it works...
- 9. Jodie vs Judith A real example ➢ Judith tends to have more life experience. ➢ Jodie was very popular for a short period compared to Judith's longevity. ➢ Neither are common in mothers groups today.
- 10. Digitally recruited Regular Givers Median Age: Actual Predicted How is it useful? In most cases, it can’t predict an individual's age with a high degree of accuracy However… It can tell us generally if one group of people are older or younger than another. It can be useful as an input to predictive models (more on that later) Age Prediction: Is it accurate? Non-Digitally recruited Regular Givers (mostly F2F) Median Age: 0 10 20 30 40 50 60 70 80 90 100 Actual Predicted 0 10 20 30 40 50 60 70 80 90 100
- 11. Using natural language processing Reviewing the free text comments on supporter motivations across different products can be insightful
- 12. ● Predictive methods can take us to the next stage of understanding things like: ○ Who to target? ○ When to target them? ○ What to target them with? ○ How much to ask for? ● We use data on supporters past interactions, demographics and publicly available data to build models that go way beyond approaches such as RFM ● We are able to measure their accuracy and be confident in our expected results Descriptive → Predictive
- 13. ● Our models score all supporters likelihood to respond to the campaign back in September. ● Results in January show our predictions were very accurate, if slightly over optimistic at the high score or very likely to donate end. ● We can predict the future! Christmas Campaign: Response
- 14. Upskilling your team. The Venn diagram is a bit outdated these days, it is from the early days of data science! However, it gives you an idea of some of the core skills. Are there people in organization who already have some of the right skills? There are plenty of pathways to learning data science today from online courses like Coursera. Classroom based courses like General Assembly or even a Masters Degree. How would I make this work for my organisation?
- 15. Is there a role for external mentoring or consultancy? Data science is the “most in-demand skill”, but what does that mean for organisations? It means difficulty hiring the right talent, high salaries and often short tenure. We are seeing this play out in the Australian market right now. External resources can bring the right skills and experience in at the time to deliver real value quickly and prove why data science is awesome! Ask me how… How would I make this work for my organisation?
- 16. Change management – be sure to bring everyone along for the journey. Data science will touch many teams within your organisation, make sure all are involved and aligned to ensure success. How would I make this work for my organisation?
- 17. Foster great communication and collaboration between the Data Scientists and Fundraisers. The crisp-DM methodology is a classic. The most important part is the repeated return to “Business understanding” or in your case “Fundraising Understanding” without this focus data science projects fall down. Make sure your teams are constantly talking to each other to ensure your data science projects are focused on the needs of your organisation. How would I make this work for my organisation?
- 18. Realistic expectations Autonomous vehicles are an extremely advanced data science / AI implementation, but they are not perfect. The question we should be asking is are they better than the existing way of doing things? The same applies in fundraising. Do not expect perfection from your first model, expect better than your current methods and a measurable outcome that delivers more fundraising dollars. Correctly set expectations and do not over sell the power of data science. How would I make this work for my organisation?
- 19. Start small but with big buy-in Pick a small quick project to start with where the outcomes are easily measurable. Share these results widely across your organisation. Build confidence as you roll out to the more complex or long term projects. How would I make this work for my organisation?
- 20. ● Start with the end in mind ● Have a clear goal, or problem to be solved ● Expect data science to be challenging, but worth the effort! How would I make this work for my organisation?
- 21. Thank you Alana Richardson Digital Marketing Manager M. 0418 791 680 E. [email protected] Li. linkedin.com/in/akrichardson James Orton Founder and Data Scientist M. 0432 795 658 E. [email protected] W. distildata.io Li. linkedin.com/in/jamesortonthedataman