Bill anatomy, order-of-magnitude cuts, engine decisions, and real-time OLAP — the corpus's deepest pool (90 companies).
From Foodpanda's −45% BigQuery bill to a 93% COUNT(DISTINCT) cut with HyperLogLog, from the Redshift→Snowflake→Databricks migration genre to Netflix's trillion-row Druid — warehouse cost and performance is the most universally interviewable material in data engineering. Includes the Data Modeling module (SCD2 mechanics, Picnic's data vault, warehouse-first architecture).
The bill decides the migration, exits run strangler-fig behind a dual-reader bridge, engine swaps follow update patterns — plus the modeling module: SCD2 mechanics and the vault that survives redefinition.
The quarterly platform review has a slide everyone dreads: two engines, one for transformation, one for BI — and a nightly copy job shuttling every processed table from the first into the second so dashboards can read it. The copy job is now a top-three line item by itself. Someone proposes consolidating; someone else says a migration would eat the year. The debate runs forty minutes on vibes until a staff engineer asks the only question that matters, and nobody can answer it: *what fraction of our spend is queries reading copies?*
The full week 3 brief is part of LeetData Pro.
