Question stated? Right chart? Right aggregation? Fair comparison? Filters obvious? Units and dates?

Six questions

A critique is not "I would have used green." It is a fixed list of questions, asked in order, each with a yes or a no and a reason. Six are enough:

1. Question stated? Is the analytic question the headline, in words, with its criterion?
2. Right chart per panel? Does each panel use the chart Module 8 prescribes for its variable type and question kind?
3. Right aggregation? Is each pivot summarising with the statistic the question asks for — Sum, Average, Count — and is it the statistic the panel title claims?
4. Fair comparison? Are compared groups on the same base — per order, per unit, same period — with the base stated?
5. Filters obvious? Can the reader see every filter and slicer that is active, on the same screen as the panels it filters?
6. Units and dates? Does every number say what it is measured in and what period it covers?

Each question is owned by a module, which is why a critique can say where a dashboard went wrong and not only that it did.

Worked example — a dashboard that fails all six

A rival crew's draft for the same Q1 data:

• No headline. The sheet is titled Dashboard v3. The reader does not know what question it answers, so cannot judge whether it does. Fails 1. Fix: "Does average order value differ by province by $100 or more? Q1 2025."
• A pie of the twelve orders. Twelve slices of amount, one per order_id. A pie composes a whole from a few parts (Module 8, Lesson 8.4); twelve slices of unrelated orders is a composition of nothing. Fails 2. Fix: a sorted bar of average by province.
• "Count of amount" by province. The pivot was built by dragging amount to Values, and Excel chose Count — ON 4, QC 3, BC 3, AB 2. The panel title says Sales by province. The numbers are order counts. Fails 3 (Module 7, Lesson 7.3). Fix: Value Field Settings → Average.
• "AB wins" on totals. Lesson 14.3's trap. Fails 4. Fix: per unit, base stated.
• A hidden filter. The monthly line shows Jan 180, Feb 90, Mar 215 — because a slicer on a sheet named Settings, which nobody opens, is set to QC. The line is correct for Quebec and labelled as the company. Fails 5. Fix: the slicer sits beside the panels it filters, and its state is visible.
• No units, no dates. The tile says 6140. Dollars? Units? Orders? Which quarter? Fails 6 (Module 2: a number without a unit is not a measurement). Fix: Total sales, CAD, Q1 2025 (Jan–Mar).

Rewritten, the draft becomes the Lesson 14.1 dashboard. That is the test of a good critique: every failure has a fix, and the fix is a lesson the crew has already had.

The peer-evaluation line

The crew project asks each member to evaluate another crew's dashboard. The form is one line per checklist question: the question, pass or fail, and — for a fail — the panel, the fault and the module that owns the fix. "3 Right aggregation — FAIL — province panel shows Count of amount, titled Sales; Module 7, Value Field Settings → Sum." Six lines, and the crew being evaluated can act on every one.

Score a second dashboard

The figure is the checklist with its owning module and classic failure; the questions hand you a fault and ask which line it fails.

CRISP-DM: the critique is evaluation → review process applied to the deployment artefact — the dashboard is checked against the question it was built to answer.