From 14-hour batches and poor documentation to AI-ready data: Mitti's dbt rebuild
Mitti cut a 14-hour dbt pipeline to 90 minutes by fixing documentation and modeling first
Overview
Mitti overhauled its data model with dbt to fix long batch cycles, poor documentation, and data-quality issues blocking AI BI adoption, achieving a 90% runtime reduction, 50%+ AI BI adoption, and $140K+ in annual savings.
What you'll take away
- 01
Full test and documentation coverage, not a new AI tool, was the actual fix: Mitti went from 40% test coverage to full coverage across 500+ models before layering AI on top.
- 02
The payoff was tangible: a 14-hour pipeline now runs in 90 minutes, recovery time dropped from 24 hours to under 2, and the rebuild cut $120,000 a year in platform costs.
- 03
Clean lineage and documentation made natural-language analytics possible: Hex became the company's fastest-adopted AI tool, growing from 30 to 150 users in its first month.
Session speakers
Yuna (Yunnan) Tang
MittiThiago Baldim
Mitti