Priya's Q3 Sales — Schema
===============================

Flat, aggregate-only teaching dataset in priya-q3-sales.csv.

Columns
-------

Column              Type     Present when                                                      Description
------------------  -------  ----------------------------------------------------------------  ---------------------------------------------------------------------------------------
dimension           text     always                                                            One of month, region, product, quarter. Identifies which rollup the row belongs to.
key                 text     always                                                            The specific label within the dimension (e.g. Jul 2026, South, Denim jackets, Q3 2026).
revenue_m           number   always                                                            Net revenue in millions of USD.
plan_or_baseline_m  number   region rows carry regional plan; quarter row carries Q2 baseline  Plan (for regions) or prior-quarter baseline (for the quarter row).
units               integer  product rows only                                                 Units sold in Q3.
gross_margin_m      number   product rows only                                                 Gross-margin dollars in millions of USD.
gross_margin_pct    number   product rows only                                                 Gross-margin as a percentage of revenue.

Empty cells are intentional: a row's dimension defines which measures apply.

Row inventory
-------------

- 3 month rows (Jul, Aug, Sep 2026).
- 4 region rows (Central, East, South, West).
- 6 product rows (Denim jackets, Leather goods, Wool sweaters, Cashmere, Cotton tees, Other).
- 1 quarter row (Q3 2026 with Q2 baseline).

Total: 14 aggregate rows.

Deliberate design choices (not imperfections)
---------------------------------------------

Unlike the CloudRevenue / M365Marketing / SupportInsights / EnterpriseGovernance teaching datasets, this dataset has no injected null-value or naming-inconsistency imperfections. Its purpose is to reproduce Chapter 1's four-panel Copilot dashboard exactly, and the dashboard defects live in the visualisation choices, not in the data.

The one designed-in tension is between raw revenue rank and plan-variance rank: South is second-smallest by raw revenue (only Central is smaller) and is uniquely and materially below plan. Ordering the four regions by raw revenue ascending puts Central first; ordering by plan variance ascending puts South first. The gap between those two orderings is what Chapter 1's regional-bar critique teaches against.

License
-------

Released for teaching use alongside The Defensible Decision. Use freely.