How Figma builds a consistent experience with confidence
Figma's data science team built an early warning system for billing bugs out of nothing more exotic than dbt tests.
Overview
Figma built an in-house testing solution combining dbt, Snowflake, and Dagster to run rigorous verification across development and production. Covers the system's architecture, design decisions, implementation challenges, and real examples where it prevented bugs from reaching users.
What you'll take away
- 01
"Consistency checkers", dbt singular tests wired into alerting, task tracking, and investigation tooling, caught billing discrepancies serious enough to have blocked Figma's IPO.
- 02
Figma now runs more than 450 consistency checkers across 15 product and business functions, catching over 100 real code and data issues in a typical month.
- 03
The checkers weren't precise out of the gate, 67% of early failures were false positives, but double-checking against a later sync and iterating with AI agents brought that down to 33%.
Session speaker
Madison Kohls
Figma