TIERRA, University of Michigan

Objective: Make years of survey data easy to explore

TIERRA is a University of Michigan research group working on health data and the use of AI in health care. Its research runs on surveys, fielded again and again, and each round arrives as its own spreadsheet. Comparing one survey to the next meant lining up questions by hand, because the same question rarely keeps the same number or wording.

TIERRA asked Web Ascender for two things: a tool to explore its survey results within and across surveys, and product concepts for a bigger idea, a national registry of the AI systems health systems use.

What Web Ascender Did

The Survey Data Analyzer

• Discovery and wireframes - a kickoff with the research team, wireframes with several ways to visualize the data, and a review and revision round before any code was written
• Upload any survey - researchers upload survey results as CSV files, and the app builds the data model behind them
• Question mapping - the same question asked in different surveys is mapped together, so results can be compared across survey waves even when the numbering changes
• Tagging - questions are tagged by topic to group and filter results
• Queries and charts - saved queries with sparklines, heat maps, pie charts, and pivot tables, with answer values sorted by strength
• Share and print - any query can be shared as a link or printed for a meeting
• Invite-only access - accounts are by invitation, so the data stays with the team
• Launch and support - production deployment, SSL, annual hosting, and ongoing fixes as new surveys bring new formats

The National Registry for AI Health (concepts)

• Exploratory wireframes for a dashboard showing which AI systems and models health systems use, refined over several rounds with the TIERRA team
• Design concepts demoed to the research team to shape the next phase of the project

The Outcome

The analyzer went from a kickoff meeting to production in about four months. Researchers now upload a new survey, map its questions to the ones they already have, and see how answers change from wave to wave, without rebuilding a spreadsheet each time.

The registry illustration is a concept and uses sample data.