Verita
Mobile app for tracking politicians’ activities and follow-through on campaign promises, empowering voters to make more informed political decisions.
Role
Database design, metric design, UI design, backend/frontend integration, data pipeline engineering
Timeline
10 Weeks
Tools
React Native, Python, Cursor, Supabase/SQL, Gemini API, Github Actions, Expo
Context
Stanford CS Capstone
[ Flagship Interactive Video Demo ]
Looping 15-second walkthrough of the finished live application
01. Customer Discovery
We interviewed 10 voters from different age demographics, testing their reactions to a low-fi prototype. We learned that voters are most interested in following their local representatives, that they found the campaign promise follow-through metric the most compelling proposed feature, and that they would be very likely to download and use the app (average 8.83 on a scale of 1-10).
"I really liked the high-level follow-through metric and voting history information, and it was great to be able to deep-dive into the specific info behind those summaries."
02. Development
Turning the prototype into a working app required intensive parallel work on sourcing and analyzing data, turning this data into meaningful metrics, and presenting these metrics together with supporting data in a simple and engaging way. We had to overcome 3 key challenges in the process: 1) We realized there is no centralized source for campaign promises—to create this data we had to build, test, and deploy AI scripts to retroactively analyze archived campaign websites. 2) We realized there is no standard categorization of legislative issues—we had to construct our own issues lexicon and then build, test, and deploy AI scripts to map campaign promises and legislation onto this structure. 3) We realized legislative activity involves huge amounts of data—we had to build, test, and deploy AI scripts to aggressively summarize legislation, meeting transcripts, and press releases to create digestible chunks of information for voters to digest.
Data Pipeline
We created Python scripts to scrape all legislative activity from the current session of the California Senate (including bills, meeting transcripts, and voting information), engineered AI prompts to analyze, categorize, and summarize this information, and designed the database backend to store it. We established a Github Actions workflow to ingest, process, and upload information to the database nightly to keep the app up-to-date with the hundreds of pages of legislative content produced every day.
Metric Development
In order to make this information digestible, we needed to create metrics that voters could understand at a glance but that were appropriately based on the detailed underlying data. We defined Action Scores to measure legislators’ activities by issue, Issue Productivity Scores to measure their relative contribution to an issue across all types of activities, and Follow-through Scores to measure their productivity on issues that they promised to act on.
User Interface
We designed the user interface to be simple and inviting, leading with high-level metrics, but always allowing users to drill down into the data behind these metrics to learn more.
[ Screenshots showing series of drilling down - followthrough score, individual issue productivity scores, detail page for issue productivity score, detail page for action ]
Intuitive high-level metrics backed by deep data drill-downs
03. Usability Testing and Product Review
We tested our high-fi prototype with users for a final round of feedback. Users reacted very positively to the app, but had several suggestions to improve data documentation and navigation clarity. Based on this feedback, we added an information page explaining how we sourced and analyzed our data and clarified some UI elements (for instance, representing links to primary sources as clearly-labeled buttons).
After addressing this feedback, the final Verita app was selected as one of the top 6 Stanford CS Capstone projects.