Скачать или смотреть Patrick Old on Surviving DBT Pipeline Chaos (Data Debug Talk)

Your DBT pipeline has hundreds of models, inconsistent naming, and someone just asked why two numbers don't match. Sound familiar?
Patrick Old (Data Lead @ Fable, previously Pave & LinkedIn) walks through 3 patterns that keep data teams from losing their minds:

1) Medallion architecture — Organize everything into bronze (raw), silver (transformed), and gold (analytics/product) layers so you always know what data to expose and where transformations belong.

2) Naming conventions — Agree on prefixes like stg_, int_, snapshot_ and write them down. No more "final_final_v2_feb2022" tables floating around.

3) Team interaction standards — Define SLAs, data contracts, and communication patterns so downstream teams know what to expect and you're not debugging 45-day-old stale data nobody mentioned.

0:00 Intro
1:15 The complexity problem
2:45 Pattern 1: Medallion architecture (bronze/silver/gold)
4:30 Pattern 2: Naming conventions that stick
7:00 Audience suggestions (the "final_final" problem)
8:45 Pattern 3: Team interaction patterns & SLAs
10:30 How to bring this to your org
11:15 Q&A

Recorded at the Jan 2026 Data Debug SF Meetup.
—
Data Debug SF is a monthly IRL meetup for data practitioners join us: https://www.meetup.com/data-debug-sf/

Proudly sponsored by Recce, your data review agent. Check us out: https://reccehq.com/