The six totals, worked fact by fact. Each one is nothing but a sum over the facts a filter accepts — no total is ever stored. The exercise is this same table with the answers blanked so you predict them first.
The answers
Every total is facts.Where(filter).Sum(Amount) — read the sum and you see exactly which facts it grouped:
| Total | Facts summed | Value |
|---|---|---|
| Row · Bonds | 400 + 410 | 810 |
| Row · Cash | 50 + 60 | 110 |
| Row · Equities | 300 + 340 | 640 |
| Column · 2024 | 50 + 400 + 300 + 200 + 100 + 10 | 1060 |
| Column · 2025 | 60 + 410 + 340 + 210 + 110 + 12 | 1142 |
| Grand total | 1060 + 1142 (or the six row totals) | 2202 |
The grand total falls out two ways — sum the columns or sum the rows — and both give 2202. That is not a coincidence; it is the cube being self-consistent, because both are the same twelve facts folded along different axes.
Live and green
The frozen battery — the same six totals with Predicted set to the truth, so every row is ✓:
Fill the six predictions with the true totals — three row totals, two column totals and the grand total — and check each against a filtered Sum over the twelve facts, so every row reads ✓ correct.
The takeaway: you never stored a single total — you derived all six from the same twelve facts. That is what "the cube is the model, the table is a view" means.