APP / PSR Regulatory Reporting · Executive Dashboard

APP Fraud Decision Evidence at a Glance

Portfolio view of APP scam decisions and evidential strength under the PSR reimbursement regime. All metrics reflect the full 494-send push-payment book across 4 episodes in this portfolio, read directly from a single engine run.

Engine run
Aizle APP fraud portfolio
aizle-consumer_APP_fraud-v9_1_0
Single run · no comparison period
Open Evidence Pack v1.3
Sends
reviewed
494
full push-payment book
Scam
sends
16
4 labelled episodes
Scam
exposure
£34,027
total customer funds at risk
Routed
Augment
494 / 494
held on the human review surface
Payments
blocked
0
corroboration bar not met
Legit sends
flagged
178 / 478
review-surface density, not an FP rate
Top reg.
distance
0.142
Romance Scam · CoP mismatch stack
Decision
confidence
0.165
uniform across the run

Decision Routing Split (494 sends)

Manual Review(Augment)
494100%
Payment Held(Protect)
00.0%
Straight-Through(Automate)
00.0%
✓No send was blocked and none was relaxed to straight-through: no send met the corroborating-evidence bar the engine requires before changing the routing action, so all evidence was routed for human augmentation.

Executive Risk Indicators

Routing posture: the 16 scam sends and the 478 legitimate sends all routed Augment; none escalated to Protect or relaxed to Automate.
Evidence differentiation: the engine differentiated evidence across scam typologies - varying detector flags and regulatory distance - without altering the routing action.
Review surface: 178 of 478 legitimate sends carried at least one flag - a density of the review surface, not a false-positive rate.
Traceability: every number on this page is transcribed from the v1.3 engine output and verified against the evidence pack, which reproduces the full decision trail.
Scope: a decision-evidence result on synthetic data - not a fraud-effectiveness or causal claim.

Scam Exposure by Typology

16scam
sends
TypologyShare of exposure*
Romance Scam38%
Investment Scam29%
Impersonation: Bank20%
Impersonation: Family12%
Total exposure: £34,027
✓*Shares rounded to the whole percent, as published in the v1.3 render - the rounding remainder sits in the smallest segment.

Evidence Stacks · Regulatory Distance

CoP mismatch + first-time payee
0.142
First-time payee + round-sum
0.117
Balance-emptying + first-time payee
0.059
First-time payee
0.029
No evidence (information only)
0.000
✓An evidence-accumulation score, not a fraud probability - the formula and severity weights are disclosed in the evidence pack (§A6) so the ranking is reproducible.

Run Integrity

✓Figures transcribed from the v1.3 engine output - and verified against the served evidence packYES
✓Scam-send decisions replayable - episode and full-book runs assert identical scam-send decisionsYES
✓Evidence pack served - full decision trail and assurance statementv1.3
▲Consumer-Duty scope - not determined by this runUNDETERMINED
▲Data ground - synthetic portfolio, no production claimSYNTHETIC

Decision Outcomes

494total
sends
OutcomeSends%
Manual Review494100%
Payment Held00.0%
Straight-Through00.0%
TOTAL494100%
✓This dashboard answers what happened; the Regulatory Evidence Pack (v1.3) answers how we know.

Executive Summary

Across 494 payment instructions, 16 APP scam sends across 4 episodes were identified with total customer exposure of £34,027. The CADi engine differentiated evidence across scam typologies - varying detector flags and regulatory distance - but did not alter the routing action: every send routed to Manual Review (Augment).

No send met the corroborating-evidence bar required before holding a payment; single-signal evidence is held for review, not blocked. Decision confidence was a uniform 0.165 across the run.

Key Findings

Scope & Limits

Where to Verify