Intelligence model
1,044 detection vectors. Calibration before conclusion.
The Mechanus IQ detection surface is 1,044 leakage vectors across 44 operational domains. Deterministic analyzers run first. The machine-learning layer runs second, and only surfaces findings after the deterministic pass has ruled out a rule-based explanation. No AI-authored finding lands on a dealer report without a named analyst review.
Model numbers
Detection vectors
1,044
across the canonical taxonomy
Operational domains
44
every dealer-facing function
Machine-learning models
24
one model per anomaly class
Risk-signal patterns
26
supply, demand, and operating-team
Deterministic first
Each vector corresponds to a declarative analyzer. Rule semantics are hand-coded by Mechanus IQ engineering and unit-tested against historical dealer exports. Analyzers are deterministic: same input, same output, every run. A finding from a deterministic analyzer never requires an explanation beyond pointing at the rule and the input row.
Machine learning layered on top
The machine-learning layer is 24 models, each scoped to a single anomaly class, calibrated against historical dealer data inside the Canadian residency boundary. No inference on client data is sent to a third-party LLM vendor outside the Mechanus IQ Bedrock routing path, which itself enforces a 0-PII contract.
Leakage taxonomy, by domain
44 domains · 1,044 vectors
D01
Sales and Desking, New Vehicles
36
D02
Sales and Desking, Used Vehicles
33
D03
F&I
49
D04
Cash Flow and Funding Velocity
36
D05
Service Department
43
D06
Parts Department
27
D07
Body Shop
23
D08
Financial Structure and Accounting
40
D09
Human Capital
41
D10
Compliance and Regulatory
37
D11
Digital Operations and Technology
30
D12
Real Estate and Facilities
21
D13
Vendor and Supplier Management
20
D14
Customer Lifecycle and Retention
26
D15
Lease and Rental Operations
16
D16
RV and Specialty Markets
20
D17
Predictive Models and ML Opportunities
32
D18
Macro and Industry Risk
23
D19
ML Anomaly Detection and Forensic Signals
76
D20
Canadian Legal Compliance
144
D21
Lender Compliance and Covenant Monitoring
39
D22
Cross-Domain Correlation Signals
48
D23
Marketing
9
D24
OEM Relationship and Audit Posture
9
D25
Reputation
8
D26
Recall and Safety Campaign Discipline
8
D27
Title
9
D28
Dealer Insurance and Risk Transfer
8
D29
Litigation
8
D30
Cybersecurity Posture and Incident Readiness
9
D31
HR Compliance
8
D32
Tax Strategy
8
D33
CapEx and Asset Management
8
D34
Wholesale and Auction Operations
8
D35
Strategic Partnerships and Joint Ventures
9
D36
Environmental Sustainability
8
D37
Trade Compliance and Cross-Border Imports
8
D38
Succession and Estate Planning
8
D39
Multi-Rooftop Coordination
9
D40
Franchise Agreement Compliance
8
D41
Brand Equity and Goodwill Tracking
8
D42
Customer Database and Marketing Automation
9
D43
Sales BDC Operations
8
D44
Chargeback Recovery Workflow
9
Discipline
Calibration
Every detector has a false positive rate target and a false negative rate target. Targets are set against historical dealer exports in the Mechanus IQ test corpus.
Explainability
Findings carry the rule that fired, the rows that triggered, and the trailing baseline used for comparison. No black-box conclusions.
Provenance
The exact ingestion run, the canonicalized row identifiers, and the analyzer build version are captured for every finding. Audit-grade by construction.
Audit posture
The canonical vector taxonomy is maintained in Mechanus IQ engineering documentation. The full inventory is available to dealers, lenders, and counsel reviewers under standard non-disclosure terms during the mini-audit phase of any engagement.