The setting

Northwind Outfitters sells camping and hiking gear online. You are an analytics engineer building the company's first modeled data layer, the trusted tables every dashboard will read from.

Every mission on this path happens at the same company, so context carries over the way it does in a real job: the data you cleaned in mission two is the data the finance lead questions in mission four.

The missions

1. source auditstarter

Audit the raw sources before Northwind's first data build

Northwind Outfitters is about to build its first modeled data layer, and the raw ecommerce tables have never been checked. Before anyone models on top of them, someone has to find out what is actually in there.

You deliver: A source audit document listing every data quality issue by table, its severity, and how to handle it downstream.

Scored on: Catches duplicate and orphan, Catches null and bad status, Catches the currency issue, Severity and handling.

Working from: raw_orders_sample.csv, source_profile.md.

2. model designstarter

Design Northwind's staging and mart layers

The raw audit is done, and now Northwind needs a real model layout: staging tables that clean the raw feeds, and marts the dashboards can query without knowing anything about the source systems.

You deliver: A model design document defining staging, intermediate, and mart models with grains, materializations, and a dependency graph.

Scored on: Clear layering, Grain clarity, Dimensions and price history, Dependency graph and materialization.

Working from: raw_catalog.md, business_questions.md.

3. transformationscore

Write the fct_orders transformation logic for Northwind

The model layout is agreed, and now the central fct_orders model needs its actual transformation logic. This is the table finance will reconcile against, so the messy raw records have to be handled correctly, not swept aside.

You deliver: A transformation specification, written as ordered SQL-style steps, that produces one clean row per order.

Scored on: Dedup to one row per order, Currency conversion, Net revenue logic, Joins and nulls.

Working from: raw_orders_txn.csv, raw_order_items.csv, fx_rates.md.

4. data testscore

Design the data tests that would have caught Northwind's bad load

A bad load slipped through to Northwind's fct_orders last month with a duplicate order, a missing customer, and a negative revenue that never should have shipped. The fix is not another manual review, it is tests that fail the build automatically.

You deliver: A data test specification mapping each test to the column it guards and the exact sample rows it catches.

Scored on: Standard tests mapped, Catches the failing rows, Custom test, No false positive.

Working from: failing_sample.csv, model_columns.md.

5. metric documentationstretch

Document Northwind's revenue metrics and close the dashboard gap

Two Northwind dashboards report different Q1 revenue, and in the last leadership meeting finance and growth spent twenty minutes arguing about which was right. Both are running off the same data, so the difference is definitional, and it is your job to settle it.

You deliver: A metrics glossary with exact definitions plus a reconciliation explaining the gap between the two revenue dashboards.

Scored on: Clear definitions, Reconciliation, Canonical choice, Edge cases.

Working from: metric_notes.md, revenue_breakdown.csv.

How the scoring works

Each deliverable is graded against the rubric written for that mission. Separately, every mission on every path is graded on how you used AI, against the same four criteria:

  • Understood the task. The learner framed the goal for the assistant clearly instead of pasting the brief and hoping.
  • Grounded in the material. The learner directed the assistant into the provided files and based the work on them, not on invented facts.
  • Verified the output. The learner checked claims, numbers, or coverage against the source material before submitting.
  • Iterated with judgment. The learner refined weak parts of the draft with specific follow-ups rather than accepting the first answer.

Both scores, with the work behind them, go on your proof profile. That is what makes a claim like "I can use AI for analytics engineering" something an employer can check.