The Challenge

Simplifying the donation sourcing experience for nonprofit customers.

Historically, when a nonprofit wanted to source donations from a donor population, they would bring in a consulting group that would raise 20% of the ask amount and use that 20% to make the next 80%. By introducing technology to the process, prioritizing which donors to pursue for a specific cause becomes quicker and less costly leading to better and cheaper outcomes for the nonprofit organization. Credera was engaged to design and build a machine-learning powered web application for nonprofits to review what donors the automated data processing pipeline. The machine learning models identify quality candidates for their fundraising campaign.

The Solution

Delivering insights by organizing and analyzing third-party data.

Credera began with an analysis of a large, semi-structured dataset it had purchased from a variety of third-party vendors. This dataset provided insight into the philanthropic giving market and how wealth and transactions were distributed within the market. Databricks provides a secure data integration capability built on top of Spark that helped immensely to unify the disparate data without centralization.The Credera team used Databricks’ Unified Analytics Engine to ingest the data, then used Databrick’s integrated workspace, designed for multi-user collaboration, to build machine learning models to determine the best potential customers of the platform, how wealth and philanthropic activity were distributed, what characteristics were common across verticals and individuals, and which individuals exhibited the best characteristics to be considered good major donors.

Once the analysis was complete, the team curated the most valuable analyses and built a Power BI Dashboard, which was connected to the Databricks Delta tables to allow for quicker and easier insights into the data.

These insights were used to inform model building, which led to a series of notebooks being developed as a pipeline to process a record from data ingestion, selection, and the machine learning scoring and finally delivered as an output to customers at scale. Databricks provides this exceptional feature of notebook workflows, which allows the organization and parameterization of multiple notebooks, which minimized the time between model development, testing, and production deployment.

The Results

Translating insights into faster identification of potential major donors.

The startup now has a high quality and responsive web application with a major data asset consisting of donor records that is enriched and assessed with machine learning models to determine the most likely donor to donate to a specific cause.

As a result of this initiative, the startup has experienced the following:

• Reduced time to identify potential major donors for clients from multiple days to just 23 minutes on average per donor.

• Enhanced customer insights by conducting usability testing. The team was able to better understand their customers and tailor their website to cater to them more effectively.

• Simplified operations for ingesting data from customer organizations to enrich and assess their donor pools to find the best potential donors for their campaign.

• Remediated bad data by creating in-depth analytics solutions to identify repeating records with different primary keys and disambiguating those records to create a more reliable and transparent pool of donors.