<DEC> What was scored

Pre-deploy: raise ASA upsell threshold from 0.71 → 0.78 on single pilot category (change_id 41A2)

Stakes

Conv. + return-rate, single category

Horizon

21-day stability window

Reversibility

Full · rollback in <30d

Domain

B2B · Operator governance

Verdict

HOLD

C clears τ by 0.01 · entropy high · queue change with named release condition

Confidence vs. threshold C 0.65 · τ 0.64 · Δ +0.01

C — confidence

0.6540

τ — threshold

0.6350

S — entropy

0.6690

One-way brake

DISENGAGED

|r_eff| = 0.654

Reasoning The strongest axis is alignment (0.78) — the change is fully reversible, scoped to one category, and sized to its risk. Evidence (0.72) and falsifiability (0.75) are also above bar: two independent return-rate sources support the thesis, and the proposed forward trigger is concrete and observable. The weakest axis is timing (0.40): the 30-day return-rate signal at the current 0.71 threshold has not completed its 21-day stability window. A threshold raise shipped against an unstable baseline cannot be cleanly attributed if returns shift downstream. C clears τ by 0.01 — a real but narrow margin. HOLD: queue the change with the named release condition (window completes est. Jun 14), then re-score. Not a refusal — a defined wait state with a clock on it.

<PRV> Provenance · per-axis confidence

evidence0.720w 0.30

differentiation0.600w 0.25

falsifiability0.750w 0.20

timing0.400w 0.15

alignment0.780w 0.10

C = Σ wᵢ·cᵢ = 0.30(0.72)+0.25(0.60)+0.20(0.75)+0.15(0.40)+0.10(0.78) = 0.654

τ = τ₀ + α·S + β·σ_stakes + γ·irr = 0.50 + 0.15(0.67) + 0.10(0.35) + 0.05(0) = 0.635

r_eff = ( +0.098, +0.049, +0.638 ) · ρ = ½ ( I + r_eff·σ ) · S(ρ) = 0.669

Instrument cluster · audit NERVA-BLOCH v11 · live readout

<QLT>Quality metrics0–1 · kernel

PURITY0.65

ENTROPY0.67

MARGIN1.03

BRAKE GAP0.88

REVERSIBLE1.00

HOMOGENEITY0.71

<MAP>Phase mapC × clarity

locus (C, 1−S)HOLD · C clears τ · S above 0.40

<FAC>Factorssigned · cᵢ − 0.5

EVIDENCE+0.22

DIFFERENTN+0.10

FALSIFIER+0.25

TIMING−0.10

ALIGNMENT+0.28

<SEN>Sensitivity+0.1 · what flips

EVIDENCE+0.030— stable

DIFFERENTN+0.025— stable

FALSIFIER+0.020— stable

TIMING+0.015— stable

ALIGNMENT+0.010— stable

C clears τ by 0.01 — narrow margin. Verdict is fragile on the downside: a 0.05 drop in evidence reverts to WAIT. No single +0.1 perturbation flips to COMMIT — S would need to fall below 0.40, requiring C ≥ 0.87. The release condition is stability data, not a confidence push.

<LOG>Agent log8 steps · 91ms

Iparsechange_id=41A2 model=nerva-bloch-v11

Iextractc_e=0.72 c_d=0.60 c_f=0.75 c_t=0.40 c_a=0.78

Ibrakereversibility=1.00 → disarmed

Iblochr=( 0.098, 0.049, 0.638 )

IcomputeC=0.654 τ=0.635 S(ρ)=0.669

IverdictHOLD · Δ=+0.019 · S above 0.40

Iauditcomponents_consistent=true

Irouteledger-append ok · release-condition=Jun-14

<RCT> Receipt — for your ledger

Action taken

Shipped Queued Escalated Blocked

Outcome (at release condition · Jun 14)

Re-scored COMMIT Re-scored WAIT N/A yet

Notes

41a27b3c92d4 Starpoint LLC · NERVA v11 · patent pending

FRAME 000091· C 0.654· τ 0.635· S 0.669· Δ +0.019· BRAKE DISENGAGED· STATE HOLD· ρ ½(I+r·σ)· r (0.098, 0.049, 0.638)· HASH 41a27b3c92d4