Back The dashboard

Deviation, measured level by level

Most dashboards score topics. This one walks the chain: activities against goals, goals against the vision, and the whole portfolio against the ethical gate. Representative composition, indicative data.

Level 1: activities and goals

Do the activities produce the outcomes?

Every objective carries an outcome indicator (EPI) and the leading behaviors expected to produce it (BES), linked by an explicit causal hypothesis. Deviation here has three named states: consistent, hypothesis under review, and outcome unexplained.

EPI, Scope 3 emissions
−7.2%
Target −8.0%, lagging outcome, correlational
BES on cadence
92%
24 of 26 leading activities on schedule
Link health
24 of 26
1 hypothesis under review, 1 outcome unexplained
Max domain PSI
0.07
Stable; bands 0.10 and 0.25 published in advance
Data sufficiency
23 of 26
3 at sentinel: insufficient data, never a score
Control C-12 effect
+0.42
CI [0.13, 0.71], causal, difference-in-differences
Level 2: goals and vision

Does what we pursue trace to the vision?

Purpose is a criterion, never a metric, so this level is a census, not a score. Either the chain from bounded context through objective to the purpose declaration exists and is documented, or the candidate has no business in the core.

Purpose chain, invariant 1
7 of 7
Objectives with a documented chain to the purpose declaration
Gate-anchored indicators
26 of 26
Each EPI traceable to a recorded gate decision
Unauthorized metrics
0
Any above zero requires retroactive anchoring or retirement
Level 3: the gate

What stands at the gate?

The gate is conjunctive: a candidate passes all five invariants or it is not admitted. There is no ethics score and no aggregate distance. What can be measured honestly: the census at the gate, headroom against published limits, and the exceptions on record.

Candidates admitted
25 of 26
1 held at invariant 3, not partially admitted
Rejected this cycle
Metric M-31
Failed invariant 2, decision recorded
Threshold headroom, invariant 3
82% and 61%
Scope 3 cap and water cap; per topic, no aggregate
Exceptions open, invariant 5
2
Oldest 41 days; every record auditor-retrievable
Externalities cycle, invariant 4
Met
All externalized effects carry named owners
Overrides this cycle
0
An override without a persisted record cannot exist
The surface

All three levels, one screen

The executive surface composes the cascade. Read-only to operators, rendered from the persisted snapshot, and every tile resolves to evidence.

npm executive
Cycle 24, DCCA snapshot, published 14:00
Invariant 1
Purpose
Invariant 2
No manipulation
Invariant 3
Limits, 1 held
Invariant 4
Externalities
Invariant 5
Auditability
Ethical gate
25 of 26 admitted
Composite, all surfaces
+1.34
+0.08 on the cycle, correlational, weights snapshot 3.2
Outcome, emissions
−7.2%
On the reduction path
Signals on cadence
92%
24 of 26, on target
Drift index
0.07
Stable, below 0.10
Scope 3, treated against control entities (causal, difference-in-differences) Internal cap NPM path, falling Control group
Held at the gate
1 indicator held at invariant 3, extraction limits
Owner: C. Díaz, day 6 of review
Rejected this cycle
Metric M-31
Failed invariant 2, decision recorded
Exceptions open
2
Oldest 41 days, records persisted
Data sufficiency
23 of 26
3 at sentinel, excluded, not scored

Evidence feed for cycle 24

14:02PassDCCA finalization: snapshot persisted, dashboard rendered, 26 indicators
13:58PassDrift check: 0.07, stable. Weight update applied, no forced recalibration
13:54PassComposite scoring: +1.34 overall, effective weights snapshot 3.2
13:48ReviewGate review: invariant 3, one indicator held on extraction limits
13:42PassStatistical validation: distribution sanity clean, no Simpson reversals
13:30PassValidation: schema and completeness pass, 26 indicators in scope
13:18PassCorrection: 1.4M rows, types, nulls and units aligned across 9 sources
06:00StartPreparation and ingestion: cycle 24 begins, sources of record loaded

Downstream of the cascade, the familiar compliance readout falls out as output, not headline:

142obligations tracked across 9 standards and regulations 3controls flagged in the security SoA F7audit programme open, 3 nonconformities, 12 observations May 4management review

Read-only to operators

Nobody edits the surface. The dashboard renders what the pipeline persisted, and nothing else.

Labeled claims

Every chart is marked correlational or causal. Missing data shows as "insufficient data", never as a score.

Evidence underneath

Every green tile resolves to an evidence chain an auditor can retrieve without asking anyone's permission.