I was inspired by the poor experience I had after donating to the Red Cross and being unable to manage my level of donations without calling them directly.

What it does

Our proposal is to create a new feature within TD My Spend that would enable individuals to manage their recurring donations online and utilize predictive analytics to prompt individuals that are likely to become new donors or increase their donations when they are algorithmically determined as likely to respond positively to such prompts.

How we built it

We used Python and SAS coding to analyze the sample data made available to us via the TD DaVinci API to develop the algorithms for analyzing customer transaction records and predicting likelihood to be a donor. We used Wix to create a mockup of how this new feature might look like and function within the TD My Spend application.

Challenges we ran into

The raw customer data which TD provided did not contain customers with donation records. Therefore, we chose to examine Netflix payments which were a recurring payment that was present in many of the sample customers. We also ran into the challenge that non-anonymized data about bank transaction would likely not be available to startups outside of the bank, so this idea is only feasible as a feature within a banking application.

Accomplishments that we are proud of

We are proud of creating a predictive algorithm with a high accuracy rate by analyzing customer bank transactions and socioeconomic attributes data.

What we learned

We learned how to navigate JSON data maps and conceptualize how data science projects could be implemented within an existing organization to enhance the quality of user experiences and serve as an intermediary for a two-sided market

What's next for OneStepUp

The TD Analytics and Data Science Leadership Development rotational program ;). In all seriousness, we would further develop our idea into a cohesive and deployable algorithm that integrates the data exploration with the predictive analysis to generate in-app recommendations

Built With

Share this project: