WHAT £7,500 BUYS · SYNTHETIC EXAMPLE · NO BUYER DATA
A FINDING THAT SHOWS
WHERE THE EVIDENCE STOPS.
The ADS Decision Risk Scorecard is Halo’s principal paid Pilot output. This synthetic proof shows how one consequential decision population becomes a bounded finding without turning uncertainty into certainty.
ONE DECISIONFOUR WEEKSSHADOW MODE£7,500 EXCL. VAT
01 / SYNTHETIC POPULATION
Concentrated decision risk can hide inside acceptable performance.
This example uses a controlled synthetic fraud-risk decision population. It is not a claim about any real organisation, buyer process or production system.
8EDGE cluster
12remainder
02 / SUPPORTED FINDING
OBSERVED SYNTHETIC RESULT
USD 3,300 of USD 4,000 observed loss occurred inside the highlighted cluster.
The controlled fixture supports an observed loss-concentration statement for this synthetic population. It does not support a claim that Halo would have prevented or recovered that loss.
LOSS CONCENTRATION82.5%observed historical association in synthetic fixture
CONTROLLED REVIEW CANDIDATES630% of eligible decisions
OBSERVED ERRORS INSIDE REVIEW SET4 / 666.7% of observed errors
03 / EVIDENCE SUFFICIENCY + PROVENANCE
Every material claim has to survive the same question: what exactly supports it?
Halo separates the visible finding from the evidence and trace that entitle it to be said. The public proof exposes the discipline, not the proprietary analysis method.
01DATASET IDENTITYLockedsynthetic dataset fingerprint retained
02LOCKED METRICcluster_observed_losspredefined analytical field
03CLAIM CLASSE2 deterministic derivedderived from supplied synthetic facts
04SUFFICIENCYLoss-finding readyminimum evidence tier satisfied
05TRACECompletecompact lineage retained
PUBLIC PROOF / PRIVATE METHOD.Raw buyer evidence, proprietary acceptance parameters, search process and internal rule ordering are not exposed.
04 / CLAIM CEILING
What this evidence supports
- A concentrated low-confidence cluster exists in the synthetic population.
- Observed errors and observed loss are disproportionately represented in that cluster.
- A bounded controlled-review population is supportable for further assessment.
- The finding can carry a compact evidence/provenance trace.
What Halo does not claim
- That every cluster decision is wrong.
- That cluster membership caused the observed loss.
- That Halo would have prevented or recovered the loss.
- That production controls should change automatically.
- That this synthetic example describes any buyer.
05 / RECOMMENDED NEXT BOUNDARY
MAXIMUM SUPPORTED MOVEMENTCONTROLLED REVIEW ONLY
Move the locked review-candidate population into controlled review for further evidence assessment before any production-control proposal.
This recommendation is evidence-bound. It is not permission to deploy, enforce, block, change a production threshold, replace a model or commence production control.
THE PAID PILOT
One consequential decision. Four weeks. One inspectable Scorecard.
The real Pilot uses buyer-approved evidence under an agreed scope and produces buyer-specific findings only where the evidence supports them.