Pinterest – Cross-Org Platform Migration
S: 2023-2024 | Staff Eng. Infra Gov. 12 platform teams each owning usage data for cost attribution (Spark platform, compute platform, KV store, S3, etc.). Old setup: each platform had bespoke SQL pipelines with heavy copy-paste. Historical onboarding took ~1 year per platform.
T: Migrate 6 platforms to the new dbt-based setup while coordinating with ~10-12 upstream teams per platform. Each migration required validating financial correctness of chargeback models (usage units × rates → cost allocation).
A1 – Standardize: Defined repeatable migration pattern: onboard platform usage data → validate chargeback model with data science → replace legacy SQL with dbt models → verify cost attribution matches.
A2 – Engage: Met each platform team to understand their data delivery contracts, cadence, and constraints. Some delivered daily, some had different schemas. Had to adapt the pattern to each platform’s specifics while keeping the canonical model consistent.
A3 – Execute: Migrated 6 platforms in 6 months with 2 engineers. Each migration involved validating that cost allocation was financially correct—wrong numbers go to C-level reporting.
A4 – Scale Pattern: Documented the migration process. New platform onboarding now follows the same playbook. Currently onboarding a platform that wasn’t even in the original implementation.
R: Per-platform onboarding ~1 year → ~1 month. 50% of system migrated. SLA breaches → 0. The repeatable pattern means any engineer on the team can now onboard a new platform.
TRAITS: Execution • Cross-Org Influence • Delivery (Q 14 16 39)