Scoring¶
Levels build on each other¶
Every level adds measures to those below (Levels build on each other:... (DEC_INCREMENTAL_LEVELS)). A stricter threshold on a higher level is a new measure there – coverage ≥ 30 % on level 2, ≥ 60 % on level 3 and ≥ 80 % on level 4 are three measures fed by the same check.
From check to area¶
Check score 0–100: binary checks give 0 or 100. Numeric checks give 100 when the threshold is met, below that linear towards 0 (
value / thresholdfor “higher is better”,threshold / valuefor “lower is better”, e.g. findings). Existence checks are binary.Measure score = mean of its checks.
Area score on a level = mean of the measures this level adds. A measure without a result (not defined, no result, broken) counts 0. Only when none of them has a result the segment shows a grey or hatched state instead – broken before defined before not defined.
Weights per measure are possible later; the first version uses the plain mean.
Reached¶
A level is reached in an area when it and every level below score 100.
The reached level of an area is the highest such level. A level above a hole can pass on its own (mutation tests on level 4 run while level 2 fails); it is shown, but not reached.
Areas are independent of each other; inside an area it is a strict ladder (Inside an area the levels a... (DEC_STRICT_LADDER)).
Project level per area = lowest level of its modules (weakest link); overall project level = lowest area level (One scale for all areas; th... (DEC_WEAKEST_LINK)).
Check types¶
Type |
Example |
Result |
|---|---|---|
automatic, numeric |
coverage, mutation score |
value + threshold |
automatic, binary |
licence rule, secrets, build starts |
pass / fail + findings |
existence |
user docs exist, threat model exists |
pass / fail; says nothing about quality |
AI-rated |
“user docs cover every feature”, brand feel of a screen |
score + reasoning + model; marked as AI rating |
human review |
pen test, legal approval |
evidence document + person + date |
Existence checks are the honest start; AI-rated and review checks deepen the same measure later.
Result states¶
State |
Meaning |
In the rings |
|---|---|---|
pass |
threshold met |
part of the score; 100 on every measure = reached |
below threshold |
not met |
score colour |
no result |
check defined, has not run, or its result is older than allowed |
dark grey |
not defined |
the model asks for the measure, the project maps no check |
light grey, dashed |
broken |
the check could not run: tool missing, crash, unreadable report |
red hatched |
waived |
deviation accepted with reason and end date |
counts as 100; the detail shows “waived until” |
regression |
was met in an earlier run, is no longer |
highlighted – the only real alarm |