Data Warehouse Performance & CostBill 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).Week 2: Order-of-magnitude optimizationsPruning is sort order, incremental models must prune the write side, sketches make distinct counts composable, and layout debt taxes every query — the mechanisms behind the 10-200× wins.

Figure 1 — where the order-of-magnitude wins live: the engine's read-decision machinery, the two big levers (scan only the new; count without scanning), layout debt, and estate discipline. By the end of this week you can predict which lever fits a given cost pathology — before running anything

The model is "incremental." It says so in the config: materialized='incremental', a tidy is_incremental() filter, fifteen million new rows a day. And every run bills like a full rebuild — thousands of slot-hours to insert a sliver of data into a 50-billion-row table. The SQL looks right. The dbt docs were followed. The bill says otherwise. Somewhere between what the query *says* and what the engine *does*, four orders of magnitude went missing. Where?After this week you canThe full week 2 brief is part of LeetData Pro.Case studies

Teej — Understanding the Snowflake Query Optimizer

Noah Kennedy — Reducing BigQuery Costs by 100–200x with dbt Incremental Models

Elad Shaabi — BigQuery HLL How we cut COUNT(DISTINCT) query costs by 93% using HyperLogLog

DoorDash — Overcoming Rapid Growth Challenges for Datasets in Snowflake

Fresha — Iceberg MoR the Hard Way: StarRocks Code Dive

Grab — Building a Spark observability product with StarRocks: Real-time and historical performance analysis.

Netflix — Optimizing data warehouse storage

Exercises

The Incremental Model That Billed Like a Full Refresh

Order-of-Magnitude Optimizations — Week 2 Check

The Full Scan That Used to Prune

To Cluster or Not — With the Bill

Materialize It, or Not — With Numbers

The Join That Made 900 Million Rows

Spilled to Remote Storage