Donation Analytics Solution for Non-Profit

Marketing & Donations

Outline

Increase total donations using a ML model

• Problem: Current mailing campaign has minimal impact on yearly donations.

• Solution: Create targeted mailing campaigns using ML

• Tools: Microsoft Excel, JMP Pro 15

Client

A Non-Profit organization that relies on donations to run their organization is seeking an effective reiteration of their standard mailing campaign. They need to create a targeted mailing campaign to engage individuals that will donate. They want to identify the attributes of their current donors to inform marketing strategy toward new potential supporters. And offer upselling to all donors. This non-profit collaborated with Tahn Analytics to design a targeted mailing campaign.

Problem: Blanket mailing campaign has minimal impact on yearly donations.

Data Analysis gauges campaign's overall success by the number of donors it elicits and the total amount of donations received. Classification and prediction models are deployed to quantify the results of the campaign. Evaluations are performed to rank and assess the best fit model. Exploratory data analysis is performed to acquire knowledge about the variables in the data set and their relationships.

Process

• Retrieve donor information from non-profit database

• Data wrangling

• Training and evaluation of Machine Learning model

Solution: Create targeted mailing campaigns using ML

To identify potential donors, we made ML models to classify and predict:

• Logistic Regression, Decision Trees, KNN, Neural Network – Models to review analytics and classification

•  Decision Trees Regression, Least Squares Regression, Neural Network – Models to preform prediction

Result: High Yield Mailing Campaign

There is an assumption that digital ads are superior to physical ads. Temple University partnered with the U.S Postal Service to produce research on print marketing and concluded that physical advertisements achieved a longer impact on participants. They took more time handling physical ads and recalled brand information faster when compared to digital ads. The research concluded that if recall and association are the targets of a marketing campaign then mailings are most effective. Our targeted mailing campaigns used ML to identify expected donors, thereby maximizing monetary contribution.