The Challenge
Vintage Curve analysis is a vital part of credit risk management.
It helps financial institutions track how loan performance degrades over time across customer cohorts.
But generating these curves manually requires deep familiarity with the data schema and the ability to write and debug SQL, often creating a bottleneck for analysts, product managers, and risk teams.
This financial institution wanted a faster, more intuitive way to get answers and insights from their loan data simply by asking questions in plain English.
A Day in the Life: Before Our Solution
An analyst wants to compare the 6-month delinquency curve for loans issued in Q1 2022 versus Q1 2023.
They have to:
- Write or reuse SQL queries
- Verify table joins and filters
- Export results to Excel
- Manually create a vintage curve graph
- Re-run everything when they change filters or cohorts
The process takes hours.
If someone less technical asks the same question — they’re stuck waiting.
Pain Points:
- Analysts had to write complex SQL for every analysis
- Vintage Curves were created manually using spreadsheets
- Business teams couldn’t self-serve or ask follow-up questions
- Changing filters or cohorts required restarting the whole process
- No integrated way to visualize credit cohort performance in real time