Every metric on the dashboard exists to answer one question. A metric without a question is decoration, and a question without a metric is a guess. The table below is the minimum set: small enough to maintain, complete enough to catch every common way RDC closure goes wrong.
Metric | Question it answers |
|---|---|
Open defects | How many findings still need fixes? The one number that must reach zero. |
Unclassified findings | How many findings still need judgement? Unclassified means open. |
Waiver count and oldest waiver age | How much closure rests on argument, and is any argument ageing? |
Blocks with clean analysis on current RTL | How much of the chip has actually been checked, on this RTL? |
Dynamic-check pass state per reset | Which resets have proven behaviour, and which only have structure? |
Coverage holes | Which required reset scenarios never ran? |
RTL or intent changes since last clean run | Is the evidence stale? Any change re-opens this row until re-check. |
Review actions open | How many review decisions are still waiting for their evidence? |
No thresholds appear in this table, and that is on purpose: what counts as an acceptable waiver age or how many blocks per week a team closes depends on the project, so the course states no numbers. Track direction and completeness instead: defects trending down, unclassified at zero, every block accounted for on current RTL. Targets Belong to the project plan, not to the methodology.
The review reads the dashboard the way it reads reports: directly, not via summaries. It checks that every metric points at current evidence, that the block list matches the chip hierarchy with nothing excluded silently, and that waiver age is measured against expiry rather than against when the waiver felt recent.
No. Unclassified findings are open findings: without a disposition nobody knows whether those twelve are defects. Classify them first.
Because a waiver is an argument with an expiry. An old waiver may rest on changed RTL or expired approval, so age signals when the argument needs renewal.