Two metrics, seven relationships.

Layer 01 — How a state is assembled

Fourteen functions per composite, and none of them recompute the source.

Each composite block reads the cumulative efficiency and profit-quality values from their own pipelines, classifies each into a spectrum class, joins the two class names into a matrix key, resolves the key into a base state and a functional role, reads the drift direction and conviction from each pipeline's interpretation table, derives the conflict type from the two drift directions, takes severity from the stronger conviction, and combines base state, conflict type, and severity into one diagnosis line. The engine references; it does not duplicate the arithmetic.

Layer 02 — Composite only

Primitive branches are excluded, deliberately.

The engine runs on the core-plan composite, the trend blend, and the all-blended view. Individual branches are left out to reduce low-frequency noise and formula load, and the reasoning is statistical rather than architectural: a single branch produces too few observations per week for a divergence between two ratios to mean anything, and a contradiction engine that fires on thin samples trains the operator to ignore it.

Layer 03 — Severity comes from conviction

The direction names the state. The magnitude decides whether to act.

Conflict type is derived from the two drift directions, and severity is taken from the conviction readings, generally the stronger of the two. This separation matters because a contradiction can be structurally real and operationally trivial. Both readings drifting apart weakly is a note; the same directions with strong conviction on either side is a finding with a review posture attached.

• Drift below half a standard deviation of structural volatility is noise, and is read as such.
• Between half and one is moderate and worth watching without acting.
• Above one and a half is a strong deviation from the metric's own structural norm.

Layer 04 — State transitions

The path between states is the leading indicator.

Any single month's relationship is a data point; the sequence is the story. Alignment decaying through transition toward contradiction is the classic early signature of a system losing grip, and it appears in this engine before it appears in blended expectancy, because the blend can stay flat while its two components move in opposite directions. The weekly history strip exists for exactly this: it compresses each week into a conflict type and severity so the trajectory is readable at a glance.