Learn Data Warehouse with Phoebe

Two tracks from dashboard pain to a working warehouse

A leader track that teaches you why warehouses exist, what they cost, and how to judge the buy landscape without writing a line of SQL - and a builder track that constructs Daybreak's first real warehouse: staging, star schema, slowly changing dimensions, loading patterns, marts, and the lakehouse. Every builder page runs DuckDB - a real columnar OLAP engine - live in your browser, with editable SQL over 300,000 rows. No install, no cloud account, no bill.

16sessions
2tracks
10live-playground sessions
300,000rows in your browser

The leader track 🤝 · for C-level, managers & curious minds · 6 x 45 min · no code to write

Thinking mode: why a warehouse is a copy on purpose, what "one truth" really costs, how to read the vendor landscape, why bills explode, and what the AI era changes - each session ends with questions to ask your data team.

🏭 Leader 1 · easy

Why a warehouse

OLTP vs OLAP in plain business terms, and the copy idea: a warehouse is your data duplicated on purpose, restructured for big questions.

▶ Leaders start here ⭐ Leader 2 · easy

The shape of a warehouse

The three layers every warehouse has, and the star schema - why facts and dimensions are the floor plan behind every dashboard.

Read-the-diagram session ⚖️ Leader 3 · medium

The one-truth problem

Why two honest teams report two different revenues: metric drift, conformed dimensions, and what "single source of truth" actually takes.

Spot-the-drift exercise 🛒 Leader 4 · medium

The buy landscape

Warehouse vs lake vs lakehouse, and the engine landscape - Snowflake, BigQuery, Redshift, Databricks - as decisions, not logos.

Vendor-map exercise 💸 Leader 5 · medium

Cost and performance

Why warehouse bills explode, the four levers that control them, and the questions that keep your team honest about spend.

Bill-autopsy exercise 🤖 Leader 6 · medium

The warehouse in the AI era

Serving ML and AI from the warehouse, why governed data is the AI moat, and a crawl-walk-run plan you can actually fund.

What-to-invest exercise

The builder track 🛠️ · for practitioners · 10 sessions x 45 min (b10: 60) · live DuckDB in your browser

One running project: Daybreak - the coffee-subscription brand from learn-sql-with-phoebe, grown up. Its OLTP database plays the source system, and session by session you build its first warehouse: staging, star schema, SCD merges, loading patterns, marts, Parquet round trips. Every page runs real DuckDB via WebAssembly - editable SQL over 300,000 rows. The engine loads once from a CDN (about 8 MB, the one network dependency); everything after that is offline.

📦 Builder 1 · easy

Row store meets column store

Feel the difference, not read about it: columnar storage vs the row-by-row app database, with DuckDB live on 300k rows.

▶ Builders start here 🏗️ Builder 2 · medium

Architecture and the staging layer

The layered warehouse, and staging as the airlock: land Daybreak's raw tables, type the text dates, and add quality gates.

Build the airlock 📐 Builder 3 · medium

Facts and dimensions

Declare the grain, split measures from context, and sketch Daybreak's star schema - the modeling move everything else hangs on.

Declare the grain 🔑 Builder 4 · medium

Building dimensions

Surrogate keys, a generated dim_date, and clean conformed dimensions - built and queried live against the staged tables.

Mint the keys 🕰️ Builder 5 · hands-on

Slowly changing dimensions

SCD Type 2 for real: customers move city, plans change price - track history and answer point-in-time questions correctly.

Time-travel joins 🔁 Builder 6 · hands-on

Loading patterns

ELT, full vs incremental loads, and the idempotent MERGE - run the same load twice and prove nothing breaks.

Run it twice 🏪 Builder 7 · hands-on

Marts and serving

Ship the warehouse to the business: department marts, views and rollups, and the finance-vs-marketing revenue question settled.

Serve the business ⚡ Builder 8 · hands-on

Performance and cost

Partition pruning, sort orders, and the cost levers - measure real scan savings live, then map them to cloud pricing.

Prune the scan 🧊 Builder 9 · hardest

Lakehouse and the modern stack

Parquet round trips live in the browser, table formats, One Big Table vs vault vs star - where the field is heading and why.

Parquet round trip 🏁 Builder 10 · hardest · 60 min

Capstone: the whole warehouse

Assemble the full pipeline end to end - raw to staging to star to marts - and close with a board pack built on your own warehouse.

End-to-end build
easy - everyone medium hands-on hardest

Best after learn-sql-with-phoebe - the Daybreak database comes from there. For schema-design depth see learn-data-modeling; for pipelines see learn-data-engineering.

Built from IBM Data Warehouse Fundamentals, DeepLearning.AI Data Engineering (Joe Reis), 365 Data Science Intro to Data Warehousing, and vendor deep dives - certificates stay official.

Choose your path 🗺️

Two doors, one warehouse. Leaders never write SQL; builders never sit through theory they have not just run themselves.

🤝 The leader path (judge & fund) a1→ a2→ a3→ a4→ a5→ a6
🛠️ The builder path (build Daybreak's) b1→ b2→ b3→ b4→ b5→ b6→ b7→ b8→ b9→ b10
🚀 The full journey (lead AND build) a1-a6→ b1-b10

The knowledge map 🧠

Both tracks at a glance - hover a session to spotlight its concepts, click any node to jump in.