27 Risk Signals: Complete Breakdown of Whistl's Impulse Detection

Whistl's Risk Orchestrator monitors 27 weighted signals to predict impulse likelihood with remarkable accuracy. From neural predictions to venue proximity, biometric vulnerability to spending velocity—this comprehensive breakdown explains every signal, its weight, and how it triggers life-saving intervention.

The Risk Orchestrator

The Risk Orchestrator is Whistl's central risk assessment engine. It combines 27 signals into a single composite score (0.0-1.0) that determines when to activate blocking, trigger AI intervention, and notify partners.

Signals are weighted based on predictive power—discovered through machine learning on thousands of user outcomes.

All 27 Risk Signals Ranked by Weight

Tier 1: Primary Predictors (>10% weight)

1. Neural Impulse Prediction (12.7%)

What It Measures: Output from the Neural Impulse Predictor—likelihood of impulse in next 2 hours

Data Sources: 56-feature input vector (time, location, biometrics, calendar, financial, behavioral, context)

Why It Matters: Most sophisticated predictor—learns your unique patterns over time

Example: Score jumps from 0.32 to 0.78 → Risk Orchestrator elevates to YELLOW

2. Spending Velocity (11.8%)

What It Measures: Rate of spending in impulse categories compared to 6-month average

Data Sources: Plaid transaction data, categorized spending

Why It Matters: Sudden spikes indicate loss of control

Example: Spending 2x normal rate in gambling category → Weight contributes 0.24 to composite

3. Neural Relapse Prediction (9.8%)

What It Measures: Likelihood of bypass/negotiation failure

Data Sources: Same 56 features plus negotiation history

Why It Matters: Predicts whether intervention will succeed

Example: High relapse score → AI deploys more aggressive negotiation steps

Tier 2: Strong Predictors (5-10% weight)

4. Venue Proximity (5.9%)

What It Measures: Physical proximity to casinos, TAB, betting shops, bars

Data Sources: GPS location, venue database

Why It Matters: Physical presence dramatically increases impulse probability

Example: Within 500m of Crown Casino → +0.059 to composite score

5. Biometric Vulnerability (5.0%)

What It Measures: HRV, sleep quality, Oura readiness score

Data Sources: Apple HealthKit, Oura Ring API

Why It Matters: Physiological state directly affects impulse control

Example: HRV 30% below baseline + poor sleep → +0.05 to composite

6. Category Spend Ratio (4.9%)

What It Measures: Current month spending vs. budget in each category

Data Sources: Connected bank accounts, user-set budgets

Why It Matters: Over-budget categories indicate失控

Example: Shopping at 150% of monthly budget → Elevated risk

7. Browsing Burst Patterns (3.9%)

What It Measures: Rapid DNS queries to shopping/gambling domains

Data Sources: VPN DNS interception logs

Why It Matters: Active browsing precedes purchases

Example: 15 gambling domain queries in 5 minutes → High risk detected

8. Calendar Proximity to Stress (3.9%)

What It Measures: Upcoming deadlines, events, known stress markers

Data Sources: Calendar integration, user-input events

Why It Matters: Stress triggers coping behaviors

Example: Major work deadline tomorrow → Risk elevated

Tier 3: Moderate Predictors (2-5% weight)

9. Sleep Deprivation (3.7%)

What It Measures: Hours slept, sleep quality score

Data Sources: Oura, Apple Health, sleep tracking apps

Why It Matters: Poor sleep impairs prefrontal cortex function

10. Emotional Distress (3.5%)

What It Measures: Self-reported mood, journal sentiment analysis

Data Sources: Mood check-ins, AI journal analysis

Why It Matters: Negative emotions drive coping spending

11. Crypto Impulse Activity (3.3%)

What It Measures: Exchange app activity, crypto domain visits

Data Sources: Screen Time API, VPN DNS logs

Why It Matters: Crypto trading shares addiction patterns with gambling

12. BNPL Stacking (3.1%)

What It Measures: Number of active Buy Now Pay Later plans

Data Sources: Transaction categorization (Afterpay, Zip, Klarna)

Why It Matters: Multiple BNPL plans indicate financial stress

13. Merchant Embedding Risk (2.9%)

What It Measures: AI-categorized similarity to known risky merchants

Data Sources: Transaction descriptions, merchant category codes

Why It Matters: New merchants similar to blocked ones pose risk

14. Time-of-Day Risk (2.7%)

What It Measures: Current hour vs. personal peak-risk hours

Data Sources: Historical impulse timestamps

Why It Matters: Late night = reduced inhibition for most users

15. Day-of-Week Risk (2.5%)

What It Measures: Current day vs. personal high-risk days

Data Sources: Historical impulse timestamps

Why It Matters: Friday nights, weekends often higher risk

16. Payday Proximity (2.3%)

What It Measures: Days since/to next paycheck

Data Sources: Plaid income detection, Argyle payroll integration

Why It Matters: Payday = fresh funds + celebration impulse

Tier 4: Contextual Predictors (1-2% weight)

17-19. Weather Conditions (1.8% combined)

Rainy days (indoor boredom → online shopping)
Cold temperatures (comfort seeking)
Seasonal affective patterns

20-21. Social Context (1.7% combined)

After social media usage (comparison triggers)
Isolation indicators (alone at home + late night)

22-23. Financial State (1.6% combined)

Balance vs. protected floor
Days until overdraft

24-25. Recent Intervention History (1.5% combined)

Blocks in last 24 hours
Bypass attempts in last 24 hours

26-27. App Engagement Patterns (1.4% combined)

Session duration spikes
Time since last check-in

How Signals Combine: Composite Score Calculation

The Risk Orchestrator calculates a weighted composite:

composite_risk = Σ(signal_value × signal_weight)

# Example calculation:
neural_prediction (0.8 × 0.127)     = 0.1016
spending_velocity (0.6 × 0.118)     = 0.0708
venue_proximity (1.0 × 0.059)       = 0.0590
biometric (0.7 × 0.050)             = 0.0350
... (23 more signals)
────────────────────────────────────────
composite_risk                      = 0.73

Risk Thresholds and Actions

Composite Score	Risk Level	SpendingShield	Actions
0.00-0.40	Low	GREEN	Normal monitoring
0.40-0.60	Elevated	YELLOW	Tighten limits, increase check-ins
0.60-0.80	High	ORANGE	Activate blocks, notify partner
0.80-1.00	Critical	RED	Full protection, crisis intervention

Signal Weight Adaptation

Weights aren't static. The Risk Orchestrator adapts based on outcomes:

Exponential Moving Average: Recent outcomes weighted more heavily
Signal Effectiveness Tracking: Signals that predict actual impulses get higher weights
Personal Calibration: Your weights differ from other users based on your patterns

Real-World Example: Complete Risk Escalation

Marcus, Friday 8:30pm, near Crown Casino:

Signal	Value	Weight	Contribution
Neural Prediction	0.72	12.7%	0.091
Venue Proximity	1.0	5.9%	0.059
Time-of-Day	0.9	2.7%	0.024
Day-of-Week	0.85	2.5%	0.021
Biometric (low HRV)	0.65	5.0%	0.033
Payday (+2 days)	0.8	2.3%	0.018
Other signals...	...	...	0.089
COMPOSITE			0.78

Action: SpendingShield goes ORANGE. VPN blocks activate. AI sends proactive alert: "You're near Crown. Your risk is elevated. Want to call your sponsor?"

Privacy: Signal Processing On-Device

All 27 signals are processed on your device:

Location data never leaves your phone
Biometric data stays in HealthKit/Oura
Transaction data processed locally
Neural network inference runs on-device

Conclusion

Whistl's 27 risk signals create a comprehensive picture of your impulse vulnerability. From neural predictions to venue proximity, biometrics to browsing patterns—every signal contributes to life-saving intervention at the moment that matters most.

Experience Intelligent Risk Detection

Whistl's 27-signal Risk Orchestrator predicts impulses before they happen. Download and let AI protect your financial future.

Download Whistl Free