Knowledge Base
How Basis Diagram Counts Are Derived
Why this page exists
The Basis diagram's numbers are checkable, and this is how to check them. If a count looks wrong, one of the rules below explains it.
Where each kind comes from
| Badge | Kind | Source |
|---|---|---|
| Pr | Uniclass product | a BOM item, grouped by its uniclass_product_code |
| Ss | Uniclass system | a BOM item's uniclass_system_code |
| ITP | Inspection and test item | an ITP item linked to the requirement |
| Ac | Programme activity | an activity linked to the requirement, or a milestone an ITP item signs off |
A requirement's section tree comes from the document's own heading structure, walked through each row's parent, so the diagram's sections are the register's sections. It does not invent groupings.
The four rules that explain almost every surprise
1. A parent's count includes collapsed children. Expanding never changes a number. If a parent reads 728 and its visible children sum to less, the remainder is in branches you have not opened.
2. One row can be counted under several requirements. A product satisfying three requirements appears under all three. This is why placements exceed rows in the header, and why summing the top-level cards can exceed the project's BOM row count. Neither is double counting: each is a distinct obligation.
3. Products with no system code are counted, not hidden. They appear in the Pr total and in incl. N with no system code. Classify them in the Uniclass document and they move under a system, without the total changing.
4. A capped read reports a floor. If the header colours source links rolled up, the server hit its row cap and every count below is a lower bound. Treat the tree as "at least this much", and do not reconcile a capped run against a BOM export.
Reconciling against the BOM
For an uncapped run, the Pr total across the whole tree equals the project's BOM item count, and the header flags a failure directly. If the tree ever fails to place a supplied row it reads MISMATCH and names the kind and the shortfall, because an aggregate that quietly loses rows is the same defect as a silent cap.
Verified on RQT-DCR: 5,970 BOM items, 67 ITP items, 44 requirements and 37 activities supplied, all placed.

The summary line above the canvas, where the reconciliation is stated: source links rolled up, products with no system code, placements, and the `MISMATCH` line when a row is lost.
If the diagram is empty
An empty tree with a non-zero link count means rows were read but none could be placed. That is a defect, not a state of your data. It happened once in development because the server projection stopped returning the fields the section walk needs (type, parent, outline) and every row was discarded silently while the page still reported itself healthy. If you see 0 nodes beside a non-zero source links rolled up, report it rather than working around it.