A real compiled decision · not a mockup

Every decision,with its receipts.

Auctorian compiles the recurring calls a business makes — pricing, markdown, replenishment — into governed decisions that carry their own evidence, policy and authority. It can't invent a number. If the evidence isn't there, it stops.

The decision on the right is real. Scroll to audit it — every figure discloses its source, and the parts we can't answer yet are shown as absent.

Auctorian Retail EngineeringPRC-Q09

Where are we leaking margin this week, and what should we do first?

Evidence11 resolved · 5 absent
current_cost
current_price
demand_velocity_trailing_4w
fx_tax_context
margin_guardrail_rules
price_elasticity_category
price_history_rolling_26w
price_ladder_category
My readRecommended

Raise DEMO-SKU-001 at DEMO-STORE-001 from $100.00 to $104.00.

Competitor at 108.00; targeting mid-market with elasticity 0.70. Margin at target 40.38%, and $155.40 of risk-adjusted margin is recoverable on this scope.

Decision quality
96
Readiness
YELLOW
Authority
HUMAN APPROVAL
Dispatch
NONE

Quality 96 alone would have permitted auto-execute. It didn't, because 5 sharpening signals are genuinely absent — so a human signs this one.

01 · Evidence

It reads what you have, and admits what it doesn't.

Every signal the decision needs is resolved, graded and attributable, or it is marked absent. Hover any value to see its tier, producer, as-of date and lineage hash.

Required · 9all 9 resolved
current_cost
HEALTHY
current_price
HEALTHY
demand_velocity_trailing_4w
HEALTHY
fx_tax_context
HEALTHY
margin_guardrail_rules
HEALTHY
price_elasticity_category
HEALTHY
price_history_rolling_26w
HEALTHY
price_ladder_category
HEALTHY
price_zone_context
HEALTHY
Sharpening · 75 absent
competitor_price_latest
HEALTHY
cross_elasticity_matrix
HEALTHY
category_foot_traffic
MISSING
kvi_flag
MISSING
promotion_calendar
MISSING
seasonality_index
MISSING
web_search_trend
MISSING
02 · Detection

It finds the work before you ask.

The margin scanner ranks scopes by recoverable value computed from observed sales, prices and costs — not by a heuristic ordering.

Opportunities · 2ranked on risk-adjusted value · USD
01
DEMO-SKU-001 · DEMO-STORE-001
underpricing leakage · 84 units observed
$155.40
02
DEMO-SKU-002 · DEMO-STORE-002
underpricing leakage · 133 units observed
$110.72

The scanner screens on a +2% scenario to rank what deserves attention. It does not price anything — the solver below does that, from the tenant's own price ladder.

03 · Deliberation

Every action competes against the alternatives.

Feasible price actions are generated from the tenant's own ladder, then scored. Doing nothing is one of them.

Candidates · 3scored by pricing_signal_candidate.v1
CONSERVATIVE
$100.000.00%
margin 38.00%
conf 0.77
0.7665
BASERecommended
$104.00+4.00%
margin 40.38%
conf 0.82
0.8157
AGGRESSIVEhighest margin · lower score
$105.84+5.84%
margin 41.42%
conf 0.72
0.7579

Every candidate is scored confidence 0.65 against margin 0.35. That is why the richest option loses: confidence carries more weight than margin, so the system declines the greediest move on its own arithmetic.

04 · Quality

The score is arithmetic, not opinion.

Decision quality

1004 = 96

Quality is not a vibe or a model confidence. It starts at 100 and every gap subtracts a named, itemised penalty. You can check the subtraction.

One of these absences costs nothing — category_foot_traffic carries a weight of zero for this decision, and the engine says so rather than rounding it up.

Penalties5 applied
seasonality_index
missing sharpening−1
web_search_trend
missing sharpening−1
category_foot_traffic
missing sharpening
promotion_calendar
missing sharpening−1
kvi_flag
missing sharpening−1
05 · Governance

Policy runs first, before the intelligence.

Authority

The strictest cap wins.

Autonomy is not a setting. Every constraint proposes the furthest it would allow this decision to go, and the most restrictive one takes it. Here two caps applied and the stricter won, so this decision routes to HUMAN APPROVAL and waits for OPERATOR.

Seeding the five absent signals would have deleted the cap that makes this approval honest. They are left absent on purpose.

Caps appliedstrongest · yellow_readiness
dqs_sovereign_candidate
AUTO EXECUTE

DQS 96.0 allows sovereign auto-execute if other caps permit

yellow_readiness
HUMAN APPROVAL

Readiness verdict is YELLOW

06 · The record

And all of it is kept.

The recordschema decision_artifact.v1
decisionacc6527dfd71bfcd
artifactartifact:7c07fd589a9d5848ba6ec2d390e930e6
signal fingerprintacc6527dfd71bfcd
registry bundle2026.05.v2 · registry_bundle_8e9636c4204413fe5b742b608e77bd8b
solver contractpricing_signal_solver.v1 · certified
certificationTHIN_VALIDATED
artifact statusDEGRADED
Where we actually are

Nothing here was dispatched.

Advisory only. Nothing on this page was dispatched to a live system, and none of it is a measured production outcome.

Dispatchnone — advisory only
Live executiondisabled
Accuracy modebacktest synthetic
Production accuracy claimnot permitted
TenantAuctorian Retail Engineering · engineering
Inputs24 signal values · 2 scopes

A generic Auctorian engineering tenant seeded with source inputs only. Not a client, and no client data is reachable from it. One complete episode has shipped. Everything beyond it is where this goes, not what it has already done.

Every figure on this page is projected from a real run · payload 0507dfb5f3f50605… · exported from d4c54cb64ea8website_decision_episode.v1