Schema & Semantic Layer

⬇ YAML ⬇ JSON

Predictive Analytics Engine

predictive-analytics v3.0

Forecast member attrition, loan defaults, deposit flight, cross-sell propensity, and growth projections with actionable interventions

🔍

Metric Definitions

at_risk_members

At-Risk Members

attrition ↓ lower

Members with high/very high attrition risk

Unit: count · Format: {value:,}

COUNT(member_id) WHERE risk_tier IN ('high', 'very_high')

predicted_attrition_30d

30-Day Predicted Attrition

attrition ↓ lower

Members predicted to leave in next 30 days

Unit: count · Format: {value:,}

COUNT(member_id) WHERE predicted_30_day_attrition > 0.5

predicted_attrition_90d

90-Day Predicted Attrition

attrition ↓ lower

Members predicted to leave in next 90 days

Unit: count · Format: {value:,}

COUNT(member_id) WHERE predicted_90_day_attrition > 0.5

attrition_model_accuracy

Model Accuracy

model performance ↑ higher

Attrition prediction accuracy

Unit: percent · Format: {value:.1f}%

correct_predictions / total_predictions * 100

Thresholds

warn low 80 target 85 stretch 90

high_risk_loans

High Risk Loans

loan risk ↓ lower

Loans with elevated default probability

Unit: count · Format: {value:,}

COUNT(loan_id) WHERE default_probability > 0.05

at_risk_loan_balance

At-Risk Loan Balance

loan risk ↓ lower

Total balance of high-risk loans

Unit: currency_millions · Format: ${value:.1f}M

SUM(principal_balance) WHERE risk_tier IN ('high', 'critical')

expected_loss

Expected Loss

loan risk ↓ lower

Projected losses from defaults

Unit: currency · Format: ${value:,.0f}

SUM(principal_balance * default_probability * lgd)

deposit_flight_risk

Deposit Flight Risk

deposit risk ↓ lower

Percentage of deposits at flight risk

Unit: percent · Format: {value:.1f}%

at_risk_balance / total_deposits * 100

Thresholds

target 3.0 warn high 4.0 crit high 5.0 stretch 2.0

at_risk_deposits

At-Risk Deposits

deposit risk ↓ lower

Deposit balance at flight risk

Unit: currency_millions · Format: ${value:.1f}M

SUM(at_risk_balance)

cross_sell_opportunities

Cross-Sell Opportunities

cross sell ↑ higher

High-propensity product opportunities

Unit: count · Format: {value:,}

COUNT(opportunity_id) WHERE propensity_score >= 0.6

cross_sell_revenue_potential

Revenue Potential

cross sell neutral

Estimated annual revenue from opportunities

Unit: currency_millions · Format: ${value:.1f}M

SUM(estimated_revenue) WHERE propensity_score >= 0.6

retention_roi

Retention ROI

retention ↑ higher

Return on retention investment

Unit: multiple · Format: {value:.1f}x

AVG(roi_multiple) WHERE outcome = 'success'

Thresholds

warn low 5.0 target 7.0 stretch 10.0

value_at_risk

Member Value at Risk

retention ↓ lower

Total LTV of at-risk members

Unit: currency_millions · Format: ${value:.1f}M

SUM(lifetime_value) WHERE risk_tier IN ('high', 'very_high')

prediction_drift_score

Prediction Drift Score

model health ↓ lower

Population Stability Index measuring shift in model prediction distributions

Unit: index

PSI(current_predictions, baseline_predictions)

Thresholds

target 0.1 warn high 0.2 crit high 0.25

feature_drift_max

Max Feature Drift

model health ↓ lower

Highest PSI across all input features indicating data distribution shift

Unit: index

MAX(PSI(feature_current, feature_baseline) for each feature)

Thresholds

target 0.1 warn high 0.2 crit high 0.3

Models & Risk Tiers

Model Performance

Attrition Model

87.3%

Accuracy

84.2%

Precision

89.1%

Recall

Last trained: 2026-01-15 · Next: 2026-02-15

Loan Default Model

92.4%

Accuracy

91.8%

Precision

93.2%

Recall

Last trained: 2026-01-10

Deposit Flight Model

85.7%

Accuracy

82.4%

Precision

88.9%

Recall

Last trained: 2026-01-12

Cross Sell Model

78.4%

Accuracy

76.2%

Precision

80.1%

Recall

Last trained: 2026-01-18

Member Risk Tiers

Very Low (0-20)

very_low

Score: 0–20

Typical attrition: 0.8%

Low (21-40)

low

Score: 21–40

Typical attrition: 2.1%

Medium (41-60)

medium

Score: 41–60

Typical attrition: 5.4%

High (61-80)

high

Score: 61–80

Typical attrition: 12.7%

Very High (81-100)

very_high

Score: 81–100

Typical attrition: 28.3%

Loan Risk Tiers

Low Risk

low

Max default prob: 0.02

Medium Risk

medium

Max default prob: 0.05

High Risk

high

Max default prob: 0.1

Critical Risk

critical

Max default prob: 1.0

Intervention Types

ID	Intervention	Cost/Member	Retention Lift	Typical ROI
early_warning	Early Warning Alert Proactive outreach at first risk indicators	$15	+23%	5.2x
risk_mitigation	Risk Mitigation Targeted intervention for medium-risk members	$45	+47%	7.1x
personalized_offers	Personalized Offers Custom retention offers based on profile	$75	+68%	8.7x
premium_service	Premium Service High-touch service for high-value at-risk members	$120	+84%	6.9x

Insight Templates

ID	Trigger Condition	Headline	Severity
attrition_spike	`at_risk_members > 2000`	Elevated Attrition Risk	warning
model_accuracy_strong	`attrition_model_accuracy >= 87`	Strong Model Performance	positive
deposit_flight_alert	`deposit_flight_risk > 3.5`	Deposit Flight Risk Elevated	warning
cross_sell_opportunity	`cross_sell_opportunities > 4000`	Significant Cross-Sell Pipeline	positive
retention_roi_strong	`retention_roi >= 8`	Strong Retention ROI	positive

Model Registry

churn_propensity

Member Churn Propensity Model

production v2.1

Algorithm: gradient_boosted_trees Training Date: 2025-11-15 Refresh: monthly Target: churn_within_90_days

Features

tenure_months transaction_frequency product_count balance_trend engagement_score

0.87

auc

0.82

precision

0.79

recall

loan_default

Loan Default Prediction Model

production v1.4

Algorithm: logistic_regression Training Date: 2025-10-01 Refresh: quarterly Target: default_within_12_months

Features

credit_score dti_ratio loan_age_months payment_history employment_length

0.91

auc

0.85

precision

0.73

recall

cross_sell_propensity

Cross-Sell Propensity Model

production v1.2

Algorithm: random_forest Training Date: 2025-12-01 Refresh: monthly Target: product_adoption_30_days

Features

product_count clv_segment tenure_months channel_preference income_estimate

0.79

auc

0.71

precision

0.68

recall

deposit_flight

Deposit Flight Risk Model

staging v0.9

Algorithm: neural_network Training Date: 2025-09-15 Refresh: monthly Target: balance_reduction_30pct

Features

balance_volatility rate_sensitivity competitor_rate_gap tenure_months

0.76

auc

0.69

precision

0.72

recall

Segment Definitions

Demographic

ID	Name / Characteristics	Filter
gen_z	Gen Z	`birth_year >= 1997`
millennials	Millennials	`birth_year BETWEEN 1981 AND 1996`
gen_x	Gen X	`birth_year BETWEEN 1965 AND 1980`
boomers	Boomers	`birth_year <= 1964`

Risk Tier

ID	Name / Characteristics	Filter
very_low	Very Low Risk	`risk_score <= 20`
low	Low Risk	`risk_score BETWEEN 21 AND 40`
medium	Medium Risk	`risk_score BETWEEN 41 AND 60`
high	High Risk	`risk_score BETWEEN 61 AND 80`
very_high	Very High Risk	`risk_score > 80`

Loan Type

ID	Name / Characteristics	Filter
auto	Auto Loans	`loan_type = 'auto'`
personal	Personal Loans	`loan_type = 'personal'`
mortgage	Mortgage	`loan_type = 'mortgage'`
heloc	HELOC	`loan_type = 'heloc'`
credit_card	Credit Card	`loan_type = 'credit_card'`

Decision Type Definitions

attrition_intervention

Attrition Intervention

retention immediate critical

Proactive retention actions for at-risk members

Triggers

⚡ at_risk_members > 1500 — At-risk member count exceeds threshold

⚡ predicted_attrition_30d > 300 — 30-day attrition forecast elevated

⚡ value_at_risk > 5M — Member value at risk exceeds threshold

Actions

• Early Warning Campaign (member_services_manager) • Risk Mitigation Program (retention_manager) • Personalized Retention Offers (marketing_manager) • Premium Service Assignment (vp_member_experience)

Outcome Metrics

at_risk_members retention_roi value_at_risk

loan_risk_response

Loan Risk Response

credit risk weekly

Actions to address elevated loan default risk

Triggers

⚡ high_risk_loans > 100 — High-risk loan count elevated

⚡ at_risk_loan_balance > 2M — At-risk loan balance exceeds threshold

⚡ expected_loss > 500000 — Expected loss projection elevated

Actions

• Enhanced Monitoring (collections_manager) • Early Collection Outreach (collections_manager) • Loan Workout Offers (lending_manager) • Loss Reserve Adjustment (cfo)

Outcome Metrics

high_risk_loans expected_loss charge_off_rate

deposit_retention

Deposit Retention

funding weekly

Actions to prevent deposit flight

Triggers

⚡ deposit_flight_risk > 3.0 — Deposit flight risk exceeds target

⚡ at_risk_deposits > 20M — At-risk deposit balance elevated

⚡ competitive_rate_gap > 50bps — Rates significantly below market

Actions

• Competitive Rate Matching (treasurer) • Relationship Outreach (retail_manager) • Loyalty Bonus Program (marketing_manager) • Product Bundling Incentives (product_manager)

Outcome Metrics

deposit_flight_risk at_risk_deposits retention_rate

cross_sell_execution

Cross-Sell Execution

growth monthly

Actions to capture high-propensity opportunities

Triggers

⚡ cross_sell_opportunities > 4000 — Significant opportunity pipeline

⚡ cross_sell_revenue_potential > 2M — High revenue potential identified

⚡ conversion_rate < target — Cross-sell conversion below target

Actions

• Targeted Marketing Campaign (marketing_manager) • Branch Referral Program (retail_manager) • Digital Recommendations (digital_manager) • Outbound Sales Campaign (sales_manager)

Outcome Metrics

cross_sell_conversion products_per_member revenue_per_member

model_optimization

Model Optimization

analytics monthly

Actions to improve predictive model performance

Triggers

⚡ attrition_model_accuracy < 85 — Model accuracy below threshold

⚡ model_drift detected — Significant model drift detected

⚡ retrain_due true — Scheduled model retraining due

Actions

• Feature Engineering Review (data_science_lead) • Model Retraining (data_science_lead) • Threshold Adjustment (analytics_manager) • New Data Source Integration (data_engineering_manager)

Outcome Metrics

model_accuracy model_precision model_recall

Decision States & Flow

stateDiagram-v2


    [*] --> surfaced




    [*] --> acknowledged




    [*] --> analyzing




    [*] --> action_planned




    [*] --> action_taken





    outcome_measured --> [*]




    deferred --> [*]




    dismissed --> [*]

Workflow

Surfaced

→

Acknowledged

→

Analyzing

→

Action Planned

→

Action Taken

→

Terminal States:

Outcome Measured (terminal)

Deferred (terminal) req. rationale

Dismissed (terminal) req. rationale

State Definitions

State	Description	Terminal
surfaced	Predictive insight identified, awaiting review	active
acknowledged	Responsible party has reviewed	active
analyzing	ROI and intervention analysis in progress	active
action_planned	Intervention strategy defined	active
action_taken	Intervention launched	active
outcome_measured	Results tracked and ROI calculated	terminal
deferred	Postponed with rationale	terminal
dismissed	Not actionable with rationale	terminal

Service Level Agreements

Decision Type	Response Window	Escalation Path
attrition_intervention	48 hours	Member Services Manager (12 hours) → Retention Manager (24 hours) → Vp Member Experience (48 hours)
loan_risk_response	72 hours	Collections Manager (24 hours) → Lending Manager (48 hours) → Clo (72 hours)
deposit_retention	7 days	Retail Manager (2 days) → Treasurer (5 days) → Cfo (7 days)
cross_sell_execution	14 days	Marketing Manager (3 days) → Sales Manager (7 days) → Cmo (14 days)
model_optimization	30 days	Data Science Lead (7 days) → Analytics Manager (14 days) → Cto (30 days)

Authority Levels

Level	Role	Can Approve
L1	analyst Predictive Analyst	— read only —
L2	member_services_manager Member Services Manager	early_warning_campaign
L3	retention_manager Retention Manager	risk_mitigation_program relationship_outreach
L3	collections_manager Collections Manager	enhanced_monitoring early_collection
L3	marketing_manager Marketing Manager	personalized_retention targeted_campaign loyalty_bonus
L3	data_science_lead Data Science Lead	feature_engineering model_retrain threshold_adjustment
L4	lending_manager Lending Manager	workout_offers
L4	treasurer Treasurer	rate_matching
L5	vp_member_experience VP Member Experience	premium_service
L5	cfo Chief Financial Officer	reserve_adjustment
L6	executive Executive Team	strategic_initiatives

Entity-Relationship Model

erDiagram
    MEMBER_RISK {
        string member_id
        integer risk_score
        enum risk_tier
        decimal predicted_30_day_attrition
        decimal predicted_90_day_attrition
        currency lifetime_value
    }
    LOAN_RISK {
        string loan_id
        decimal default_probability
        enum risk_tier
    }
    DEPOSIT_RISK {
        string account_id
        integer flight_risk_score
        currency at_risk_balance
    }
    CROSS_SELL {
        string opportunity_id
        decimal propensity_score
        currency estimated_revenue
    }

Entity Details

Member Risk Profile

Attrition probability and risk scoring for members

6 attrs

Attribute	Type	Key
member_id	string
risk_score	integer
risk_tier	enum
predicted_30_day_attrition	decimal
predicted_90_day_attrition	decimal
lifetime_value	currency

Loan Risk Assessment

Default probability for loan portfolio

3 attrs

Attribute	Type	Key
loan_id	string
default_probability	decimal
risk_tier	enum

Deposit Flight Risk

Flight risk scoring for deposit accounts

3 attrs

Attribute	Type	Key
account_id	string
flight_risk_score	integer
at_risk_balance	currency

Cross-Sell Opportunity

Product propensity scoring

3 attrs

Attribute	Type	Key
opportunity_id	string
propensity_score	decimal
estimated_revenue	currency