ModernCS
Join for free
Snowflake

Data · Taught with Snowflake

Big Data / Analytics

Analytical SQL, dimensional modelling and dbt. You build tested, documented models that other teams can query without asking you first, and you keep the warehouse bill sane.

Download the course plan
  • Foundational
  • 4 phases
  • 21 sessions
  • First session free

Syllabus

4 phases · 21 sessions · each phase ends in something you have built

1

Questions and Answers

Answer a real business question in SQL without asking anyone for a cloud account.

$72

this phase

  1. 1.1

    What Analytics Owns

    Free preview

    The job in one sentence: a question in, a number someone acts on out, and why that is not the job of moving the data.

    Open
  2. 1.2

    Getting Data In

    Load CSV and Parquet, fix the column that arrived as text, and look at a dataset before trusting it.

  3. 1.3

    SQL That Answers Something

    SELECT, WHERE, GROUP BY, and reading your own result set with suspicion.

  4. 1.4

    Joins and the Fan-Out

    How one extra join doubles revenue, and how to catch it before anyone else does.

  5. 1.5

    Windows and Cohorts

    CTEs, window functions, retention, and building a hard query one readable step at a time.

By the end of this phase:One SQL file that answers six questions about a raw dataset, run locally at zero cost.
2

From Queries to Models

Turn a folder of queries into a project someone else can run.

$72

this phase

  1. 2.1

    Dimensional Modeling

    Kimball's four steps: pick the process, declare the grain, then facts and dimensions.

  2. 2.2

    Your First dbt Project

    Sources, refs, models, and a build order you never had to draw yourself.

  3. 2.3

    Staging, Intermediate, Marts

    The three layers, the naming rules, and why a 400-line query is not a model.

  4. 2.4

    Materializations

    View, table, and ephemeral: trading build time against query time on purpose.

  5. 2.5

    Tests and Freshness

    not_null, unique, relationships, and catching a broken source before anyone opens the dashboard.

By the end of this phase:A layered dbt project on DuckDB that builds and tests itself from one command.
3

Off the Laptop

Get the project onto a real warehouse and a real schedule without breaking it.

$72

this phase

  1. 3.1

    Git for One Person

    Branches, commits, pull requests, and getting yesterday's version back.

  2. 3.2

    Into Snowflake

    Accounts, virtual warehouses, storage and compute apart, micro-partitions, and a 30-day trial clock you start on purpose.

  3. 3.3

    Same Project, Two Targets

    Develop on DuckDB, deploy to Snowflake, and change one profile instead of every model.

  4. 3.4

    Incremental Models and Snapshots

    Rebuild only what changed, and keep SCD Type 2 history that does not quietly rewrite itself.

  5. 3.5

    Docs and Lineage

    Generated documentation, column-level lineage, and answering "where does this number come from" in ten seconds.

  6. 3.6

    Deploy: CI and Schedules

    Checks on every pull request, a nightly production run, and an alert when it fails at 3am.

By the end of this phase:The same project running nightly on Snowflake, with tests gating every pull request.
4

Trust, Cost, Consumers

Make the numbers agreed on, affordable, and actually used by someone.

$72

this phase

  1. 4.1

    Metrics and the Semantic Layer

    One definition of "active user", written once in YAML and reused by every tool that asks.

  2. 4.2

    Questions Without SQL

    Semantic views, Cortex Analyst, and MCP: a natural-language answer is only as good as the model underneath it.

  3. 4.3

    Warehouse Cost

    Credits, auto-suspend, adaptive warehouses, budgets, and a resource monitor that stops the bleeding.

  4. 4.4

    Access and PII

    Roles, grants, masking policies, row access, and a data request you are allowed to refuse.

  5. 4.5

    Delivering Analytics

    Dashboards as code, self-serve for people who will not write SQL, and the readout where someone disagrees with your number.

By the end of this phase:A published metric with one named owner, one dashboard, and a monthly cost you can state.

Tools you will use

  • Snowflake
  • dbt Fusion
  • DuckDB
  • Parquet
  • MetricFlow
  • Apache Ossie
  • Evidence
  • GitHub Actions

What you will build

  • Six Questions, One SQL File

    A raw dataset answered on your laptop, zero cloud spend

  • Two Warehouses, One Project

    Develop on DuckDB, ship to Snowflake, change one file

  • The Metric Nobody Argues About

    One definition, one owner, one number