Communicate• Understand your audience• Practical analytics not data science• Easy to go too far
What is a predictive model? Find those that look like your donors andyou will have a better chance of producing more donors!• Gather data about your constituents• Find data with predictive power• Combine data to produce a model
What gives data predictive power? What does the average donor look like?• Predictive models use distinguishing characteristics not common characteristics• Do not look only for similarities between your donors• Look for distinguishing qualities between your donors and the rest of your constituents
The answers…Email address = COMMON characteristicLegacy pledge = DISTINGUISHING characteristicMAJORITY of donors have email yet MINORITY ofthose with email are donors.MINORITY of donors have pledged legacy yetMAJORITY of legacy pledgers are donors.
The question is NOT “Why do people give?”.xkcd.com
Selecting VariablesGiving history AgeWealth indicators Questionnaire/Survey responderInterests Email clicksAffiliations Twitter/facebookGender Events attendedSign up/subscriptions Family relationshipsEmployment/positions AddressMarital status EmailDegree PhoneMailing preference (opt outs) First gift amountVolunteers Proximity
Prepare your data file Constituent Is a donor? Attended Has email? Over 40? ID Event? A 1 1 1 1 B 1 0 1 1 C 0 1 1 0 D 1 1 0 1 E 0 0 1 1• Excel v SPSS
Conclusions….• The average donor and the average non-donor may look the same.• Look for distinguishing characteristics not common ones.• Don’t look at donors in isolation. Compare data for donors with data for everyone.
Conclusions….• Data modelling can help you focus your resources on the best prospects.• Demonstrate worth on low risk segments.• Consider your audience. Communicate results so that everyone can understand.
Paul WeighandInsight ManagerUniversity of Edinburgh@paulweighand