**The brief.** A total in a cube is never *stored* — it is *computed* by summing every fact at a coordinate. Six asset positions, two years, twelve facts. **Predict six totals, then press Run.** Every ✗ is a place your arithmetic and the cube disagree — correct the prediction until all six say ✓.

This is Lesson 1 in miniature: the flat list of facts is the model, and every row total, column total, and grand total is just a different grouping of those same twelve numbers. Nothing here is ever entered twice.

The rules

• The twelve facts are fixed, and so is each total's definition (the Filter that says which facts it sums). Only the Predicted numbers move.
• The Predicted column is pre-filled — some are right, most are wrong.
• Run, read the ✗ rows, and correct each Predicted until all six say ✓. You have then predicted the cube by hand.

No peeking at the solution until you have been wrong at least once — the gap between your guess and the fact is the lesson.

Your workbench

When every row says ✓, you have proved the point of Lesson 1 by hand: the totals were never separate data — they were always just the facts, regrouped.

Wrong at least once? Reveal the solution — the same six totals with the fact-by-fact sums behind each number.

← Lesson 1 exercises · Lesson 1 — Why model as a cube