🏛️ Data Warehousing

OLTP vs. OLAP, facts and grain, columnar engines (DuckDB and ClickHouse), cubes and pre-aggregation.

No lessons published yet — this is the roadmap, so you can see exactly what the course will cover before it lands.

Before you start: Learn SQL — Series 2. Everything here is a GROUP BY over a join, then the same query made to read less data. Series 2 is where multi-table joins and CTEs stop being an obstacle.

Short on time? The 23 episodes marked core are the shortest path to being productive at work. Start there and pick up the rest whenever a job asks for them.

Series 1 · Warehousing Fundamentals

build a warehouse, load it, and prove it matches the source

01
OLTP vs OLAP — Two Workloads That Need Different Table Designs
corebeginner#kimballcoming soon
02
Warehouse, Lake & Lakehouse — What Each One Actually Is
corebeginner#kimball#inmoncoming soon
03
The Star Schema from the Query Side — Why Analysts Like It
corebeginner#kimballcoming soon
04
Facts, Dimensions & Grain in Practice — Reading Someone Else's Warehouse
corebeginner#kimballcoming soon
05
Loading the Warehouse — From Northwind OLTP to fact_sales
corebeginner#kimballcoming soon
06
Reconciling With the Source — Proving the Warehouse Agrees With the App
corebeginnercoming soon
07
Slice, Dice, Roll-Up & Drill-Down — The Four Analytical Moves
corebeginner#kimballcoming soon
08
ROLLUP, CUBE & GROUPING SETS — Subtotals the Standard Way
corebeginnercoming soon
09
Pre-Aggregation — Summary Tables and Materialized Views
corebeginner#kimballcoming soon
10
Time-Based Analysis — The Date Dimension Doing Real Work
corebeginner#kimballcoming soon
11
Project — One Star Schema, Loaded and Queried
corebeginner#projectcoming soon

Series 2 · Columnar Engines & Cubes

make queries fast with columnar engines and cubes

01
Columnar Storage — Why Analytics Loves Columns
coreintermediate#stonebrakercoming soon
02
Inside a Parquet File — Row Groups, Encodings and Why Skipping Works
coreintermediate#melnik-dremelcoming soon
03
DuckDB — An Analytical Warehouse That Runs Inside Your Process
coreintermediatecoming soon
04
ClickHouse & MergeTree — Columnar Storage on a Server, at Scale
coreintermediatecoming soon
05
Partitioning, Sort Keys & Data Skipping — Read Less, Answer Sooner
coreintermediatecoming soon
06
Storage and Compute, Separated — Reads Cross a Network, Compute Can Sleep
coreintermediatecoming soon
07
The Semantic Layer — Dimensions, Measures and Hierarchies, Declared Once
coreintermediatecoming soon
08
Kinds of Semantic Layer — In the BI Tool, Standalone, or in Your Project
intermediatecoming soon
09
Reading Cubes Somebody Else Built — Sales, Inventory, Customers and HR
intermediatecoming soon
10
Querying a Cube — Pick the Fields, Then Read the SQL It Wrote for You
intermediatecoming soon
11
Materialized Views & Incremental Aggregation — Compute Once, Read Often
coreintermediatecoming soon
12
Star vs One Big Table on Columnar Engines — Measured on the Same Data
intermediatecoming soon
13
Compression & Encoding — Why the Same Table Is Ten Times Smaller Here
intermediatecoming soon
14
Project — The Same Question, Reading a Tenth of the Data
coreintermediate#projectcoming soon

Series 3 · Advanced Warehousing

billions of rows, streaming loads, and controlling the cost

01
Query Performance on Columnar Engines — Reading the Plan Again
coreadvancedcoming soon
02
Aggregating Materialized Views & Projections — ClickHouse Computing the Total at Insert Time
advancedcoming soon
03
Slowly Changing Dimensions in the Warehouse — Querying History Correctly
coreadvanced#kimballcoming soon
04
Billions of Rows — Partitions, Sharding & Distribution
advancedcoming soon
05
Lakehouse — Parquet, Iceberg and DuckDB Straight on Files
advanced#armbrust-lakehousecoming soon
06
Querying Across Databases — Federation, and Where the Join Actually Runs
advancedcoming soon
07
Streaming Ingestion — Minutes Behind Instead of a Day
advancedcoming soon
08
Cost & Storage — Compression, TTL and Tiering
advancedcoming soon
09
Benchmark — The Same Star Schema on DuckDB vs ClickHouse
advancedcoming soon
10
From Warehouse to BI — Feeding Dashboards, Cubes and Extracts
advancedcoming soon
11
Project — The Whole Northwind Warehouse, on Two Engines, Reconciled
coreadvanced#projectcoming soon

Where this leads

Apache Iceberg & Table Formats — starts from Series 3 · 20.
Trino — starts from Series 2, and Series 3.

Not here, on purpose

Designing the dimensional model — grain, SCD type, star vs snowflake — covered in Data Modeling; here you build and query the model
The pipelines that load it — covered in ETL & Data Pipelines and Analytics Engineering with dbt
Dashboards, metric definitions and delivery to business users — covered in BI & Data Visualization
Writing the analytical queries themselves — window functions, CTEs, tuning — covered in Learn SQL — here the query is given and the question is how much data it had to read
Running the server — backups, replication, upgrades, on-call — covered in Data Ops
Reading a cube from a dashboard, and the charts built on it — covered in BI & Data Visualization, which consumes a cube. In this course, Series 2 designs and builds one
Deciding the star's grain, its SCD type and star vs snowflake — covered in Data Modeling, Series 3. In this course, Series 1 · 20 loads a design somebody has already made and Series 1 · 10 queries it

🏛️ Data Warehousing

Series 1 · Warehousing Fundamentals

OLTP vs OLAP — Two Workloads That Need Different Table Designs

Warehouse, Lake & Lakehouse — What Each One Actually Is

The Star Schema from the Query Side — Why Analysts Like It

Facts, Dimensions & Grain in Practice — Reading Someone Else's Warehouse

Loading the Warehouse — From Northwind OLTP to fact_sales

Reconciling With the Source — Proving the Warehouse Agrees With the App

Slice, Dice, Roll-Up & Drill-Down — The Four Analytical Moves

ROLLUP, CUBE & GROUPING SETS — Subtotals the Standard Way

Pre-Aggregation — Summary Tables and Materialized Views

Time-Based Analysis — The Date Dimension Doing Real Work

Project — One Star Schema, Loaded and Queried

Series 2 · Columnar Engines & Cubes

Columnar Storage — Why Analytics Loves Columns

Inside a Parquet File — Row Groups, Encodings and Why Skipping Works

DuckDB — An Analytical Warehouse That Runs Inside Your Process

ClickHouse & MergeTree — Columnar Storage on a Server, at Scale

Partitioning, Sort Keys & Data Skipping — Read Less, Answer Sooner

Storage and Compute, Separated — Reads Cross a Network, Compute Can Sleep

The Semantic Layer — Dimensions, Measures and Hierarchies, Declared Once

Kinds of Semantic Layer — In the BI Tool, Standalone, or in Your Project

Reading Cubes Somebody Else Built — Sales, Inventory, Customers and HR

Querying a Cube — Pick the Fields, Then Read the SQL It Wrote for You

Materialized Views & Incremental Aggregation — Compute Once, Read Often

Star vs One Big Table on Columnar Engines — Measured on the Same Data

Compression & Encoding — Why the Same Table Is Ten Times Smaller Here

Project — The Same Question, Reading a Tenth of the Data

Series 3 · Advanced Warehousing

Query Performance on Columnar Engines — Reading the Plan Again

Aggregating Materialized Views & Projections — ClickHouse Computing the Total at Insert Time

Slowly Changing Dimensions in the Warehouse — Querying History Correctly

Billions of Rows — Partitions, Sharding & Distribution

Lakehouse — Parquet, Iceberg and DuckDB Straight on Files

Querying Across Databases — Federation, and Where the Join Actually Runs

Streaming Ingestion — Minutes Behind Instead of a Day

Cost & Storage — Compression, TTL and Tiering

Benchmark — The Same Star Schema on DuckDB vs ClickHouse

From Warehouse to BI — Feeding Dashboards, Cubes and Extracts

Project — The Whole Northwind Warehouse, on Two Engines, Reconciled

Where this leads

Not here, on purpose