◆

Small Business — Industry-Contextual Analysis

Understanding business nature through expected vs. actual patterns, industry benchmarks, and network flow analysis. Different industries have fundamentally different "normal."

Research Note: Small Business AML is Under-Studied

Unlike retail AML (well-established typologies), small business patterns vary dramatically by industry. This module builds industry-specific baselines for expected transactions (sales, payroll, supplier payments, tax remittances) and flags deviations from those expectations. The approach is inherently contextual — a cash-heavy restaurant looks very different from a consulting firm.

GD
Golden Dragon Restaurant
Small Business · SB-2024-03312 · Incorporated: 2017-06-22
Food Service (NAICS 7225)
Richmond, BC
$850K/yr
12 FTE
Dine-in + Takeout
$72K (last 6m)
68% ▲
14%
Composite Score
0.88
Industry-adjusted anomaly
across revenue + payroll

Contextual Anomaly Score

CWAE-MMD
0.91
97th percentile among Food-Service businesses in BC
Context-Adjusted Percentiles (behavior | industry + size)
Cash Deposit Ratio
P92
Payroll / Revenue
P3 ▼
Personal Outflows
P89
Revenue Volatility
P85
Tax Regularity
P78
Supplier Diversity
P42
COGS Ratio
P48
Interpretation: Given the business profile (food service, 12 FTE, $850K revenue, Richmond BC), the model expected ~40% cash and ~30% payroll. Observed cash at 68% and payroll at 14% are extreme outliers. The reconstruction error is dominated by these two features.

SHAP Feature Importance

SHAP
SHAP values — impact on anomaly score prediction. Red = pushes score higher, blue = pushes lower.
payroll_to_revenue
+0.182
cash_deposit_ratio
+0.163
personal_outflow_pct
+0.108
revenue_seasonality
+0.087
tax_remit_regularity
+0.065
unidentified_depositors
+0.048
cogs_ratio
-0.038
supplier_consistency
-0.029
rent_ratio
-0.018
account_age
-0.012
Base value: 0.32 (population mean) Model output: 0.88 (Δ = +0.56)
vs. DynamicLRP: SHAP and LRP agree on top-2 drivers (payroll, cash). SHAP additionally highlights revenue_seasonality (flat revenue = manufactured). LRP provides conservation guarantee; SHAP provides additive decomposition from base rate.

Expected vs. Actual Business Activity

FOOD SERVICE BASELINE
Industry baseline: Full-service restaurant, Richmond BC, 12 FTE, $850K declared revenue
Business Function Expected Range Actual (3m avg) Deviation Signal
Sales Revenue (deposits) $55K–$80K/mo $72K/mo Normal Within range, but see cash ratio below
Cash Deposits 35–50% of revenue 68% ($49K/mo) +36% above Excessive cash for modern restaurant
Payroll 28–35% of revenue 14% ($10K/mo) -50% below Impossibly low for 12 FTE
Supplier / COGS 28–35% of revenue 31% ($22K/mo) Normal Food cost within range
Rent / Lease 6–10% of revenue 8% ($5.8K/mo) Normal Consistent with Richmond commercial
Utilities 3–5% of revenue 4% ($2.9K/mo) Normal Normal range
Tax Remittances Regular (GST/PST monthly) Irregular Irregular Skipped 2 of last 6 months
Unexplained Outflows ~0% $8,200/mo to personal accts Anomaly Cash out to 3 personal accounts

Industry Benchmark Comparison

CWAE-MMD
Green range = industry P25–P75. Marker = this business.
Cash Ratio
68% (P92)
Payroll / Rev
14% (P3)
COGS / Rev
31% (P48)
Rent / Rev
8% (P45)
Rev / Employee
$71K (P72)
Deposit Frequency
18/mo (P55)
Key anomaly: Cash ratio at P92 and payroll at P3 — the business claims 12 employees but payroll is half what the cheapest restaurant in the peer group pays. Either employees are unreported (cash-in-hand) or the headcount is fabricated.

Business Money Flow

FLOW ANALYSIS
INFLOWS
Cash Sales
$49K/mo
Card/Digital
$23K/mo
Golden Dragon
$72K/mo
→→
→
OUTFLOWS
Suppliers
$22K
Payroll ⚠
$10K
Rent
$5.8K
Utilities
$2.9K
Tax ⚠
Irregular
Personal Accts ⚠
$8.2K
→
→
→
Unaccounted gap: ~$23K/mo (32% of revenue)

Feature Attribution (Business Context)

DynamicLRP
Signed relevance — conditioned on industry: Food Service
payroll_to_revenue
+0.38
cash_deposit_ratio
+0.31
personal_outflow
+0.22
tax_regularity
+0.11
supplier_consistency
-0.04
revenue_trend
-0.03
deposit_frequency
-0.02
Interpretation: The anomaly is driven almost entirely by the payroll-revenue disconnect and excessive cash ratio. Supplier costs and revenue levels are normal — the business looks legitimate on the revenue side, but the cost structure is impossible.

Revenue Seasonality vs. Industry

PATTERN
Monthly revenue — actual vs. food-service industry seasonal pattern
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
■ Industry avg ■ Actual
Suspiciously flat — no seasonal variation
Pattern: Food-service businesses typically show 15–25% seasonal variation (summer peaks, Jan dip). Golden Dragon shows <5% variance — consistent with manufactured revenue rather than genuine customer-driven sales.

AML Typology Mapping

COMPOSITE
Typology match scores (multi-lens consensus — adapted for small business)
Placement
0.86
High Match
Cash-Intensive
Business Front
0.82
High Match
Structuring
0.61
Possible
Payroll
Manipulation
0.74
Moderate
Tax Evasion
0.58
Possible
Layering
0.29
Unlikely
FINTRAC Special Projects
Underground Banking
4 of 7 red flags matched
● Cash-intensive business, unidentified depositors, personal commingling
Fentanyl
1 of 9 red flags matched
○ Cash-intensive business in known corridor
Trade-Based ML
0 of 6 red flags matched
○ No match
Human Trafficking
0 of 8 red flags matched
○ No match

Red Flag Indicators

RFIL
  • ■
    Payroll impossibly low for declared headcount — $833/employee/mo vs BC minimum wage ~$3,016/mo FT
    Source: CWAE-MMD + Industry Baseline · Confidence: High
  • ■
    Cash deposit ratio (68%) exceeds P92 of food-service peers — inconsistent with modern restaurant in metro area
    Source: SHAP + LRP Attribution · Confidence: High
  • ■
    $8,200/mo flowing to 3 personal accounts of owner and family — business/personal commingling
    Source: Flow Analysis · Confidence: High
  • ▲
    Three unidentified third-party depositors (C1-C3) making regular cash deposits on behalf of the business
    Source: GNN Network · Confidence: Medium
  • ▲
    Revenue shows <5% seasonal variation — food-service peers show 15–25%. Consistent with manufactured revenue.
    Source: Seasonality Analysis · Confidence: Medium
  • ○
    Irregular GST/PST remittances — skipped 2 of last 6 months
    Source: Expected vs Actual · Confidence: Low
  • ○
    Unaccounted gap of ~$23K/mo (32% of revenue) between inflows and identified outflows
    Source: Flow Analysis · Confidence: Low

Transaction Network Pattern

GNN
S1
S2
S3
GD
PERS
C1
C2
C3
Suppliers (normal)
Cash sources (?)
→ Personal outflow
Subgraph: 8 nodes · 10 edges · 2-hop ego
Pattern: FAN-IN + FUNNEL-OUT (conf: 0.84)
Business
Supplier (outflow)
Cash source (inflow)
Personal (outflow)
Network Pattern: FAN-IN from 3 unidentified cash depositors (C1-C3) + FUNNEL-OUT to personal accounts. Legitimate supplier relationships (S1-S3: food wholesaler, beverage distributor, equipment) are normal. The combination of external cash injection and personal extraction is consistent with a cash-intensive business front typology.

Investigation Narrative

LLM-CoT
Step 1: Business Profile
Golden Dragon Restaurant (SB-2024-03312) is a full-service restaurant in Richmond, BC with 12 declared employees and $850K annual revenue. The business has been operating for 8 years with a stable supplier base.
Step 2: Anomaly Analysis
Two critical anomalies emerge from the industry-contextual analysis:
(a) Payroll impossibility: At 14% of revenue ($10K/mo for 12 FTE), payroll works out to ~$833/employee/month — well below minimum wage in BC ($17.40/hr = ~$3,016/mo FT). Either the employee count is fabricated or workers are paid cash off-books.
(b) Cash ratio excess: At 68% cash, this exceeds the P92 of food-service peers. Modern restaurants in metro Vancouver typically see 35-50% cash. Combined with unidentified cash depositors (C1-C3), this suggests possible cash injection from external sources.
Step 3: Hypothesis
Two plausible scenarios: (1) Cash-based laundering: External illicit cash is deposited through the restaurant as "sales revenue," inflating the cash ratio while keeping other business metrics normal. The low payroll suggests employees are also paid in cash to reduce paper trail. (2) Tax evasion with off-books payroll: Revenue is genuine but employees are paid cash to avoid payroll taxes/CPP/EI, with excess profits extracted to personal accounts.
Step 4: Recommendation
Escalate for L2. Key actions: (1) Identify cash depositors C1-C3 — are they employees or third parties? (2) Cross-reference employee list with CRA records. (3) Request POS data to validate cash-vs-card ratio. (4) Review personal account activity of owner. Consider STR for potential placement if external cash injection is confirmed.

Add-On Model Signals

PLUG-IN
POS Integration
—
Not connected
CRA Cross-Ref
—
Not connected
TM Rule Engine
—
Not connected
Industry Model
+
Add plug-in