metadata
license: apache-2.0
task_categories:
- tabular-classification
- tabular-regression
tags:
- nigeria
- banking
- churn-prediction
- customer-analytics
- customer-360
- fintech
- customer-segmentation
language:
- en
size_categories:
- 1M<n<10M
Nigerian Banking Customer 360
Dataset Type: Banking & Finance - Customer Analytics
Version: 1.0
License: Apache 2.0
Language: English
Geography: Nigeria
Dataset Description
Synthetic Nigerian banking customer 360-degree profiles with churn prediction labels. Comprehensive view of customer demographics, product holdings, balances, transaction activity, and digital engagement.
Use Cases:
- Customer churn prediction
- Lifetime value (LTV) modeling
- Next-best-offer recommendations
- Customer segmentation
- Product cross-sell/up-sell
- Retention campaign targeting
Dataset Statistics
- Rows: 2,000,000 customers (pilot: 10,000)
- Columns: 24
- Snapshot Date: 2024-12-31 (monthly snapshots possible)
- Churn 30d Prevalence: 5%
- Churn 90d Prevalence: 12%
Schema
| Column | Type | Description |
|---|---|---|
customer_id |
string | Unique customer identifier (CUS-########) |
snapshot_date |
date | Data snapshot date |
age |
int | Customer age (18-70) |
gender |
category | male, female |
monthly_income_ngn |
float | Monthly income in Naira |
state |
category | Nigerian state (37 states) |
kyc_tier |
category | tier1, tier2, tier3 |
account_age_days |
int | Days since account opening |
products_count |
int | Number of products held (1-7) |
has_savings_account |
bool | Has savings account |
has_current_account |
bool | Has current account |
has_loan |
bool | Has active loan |
has_debit_card |
bool | Has debit card |
has_credit_card |
bool | Has credit card |
has_wallet |
bool | Has digital wallet |
has_investment |
bool | Has investment product |
total_balance_ngn |
float | Total balance across products |
transaction_count_30d |
int | Transaction count (last 30 days) |
transaction_volume_ngn_30d |
float | Transaction volume (last 30 days) |
days_since_last_transaction |
int | Days since last transaction |
dormant_flag |
bool | Dormant if no activity in 90+ days |
digital_engagement_score |
int | Digital engagement score (0-100) |
churn_30d |
bool | LABEL: True if churned in 30 days |
churn_90d |
bool | LABEL: True if churned in 90 days |
Label Distribution
Churn 30d
- Positive: 5%
- Negative: 95%
Churn 90d
- Positive: 12%
- Negative: 88%
Note: churn_90d is a superset of churn_30d
Churn Drivers
High Risk Factors
- Dormant (90+ days no activity): 15x risk
- Single Product: 5x risk
- Low Engagement (<30 score): 6x risk
- Low Balance (<₦10k): 4x risk
- Failed Transactions (>5/month): 3x risk
- No Direct Deposit: 3.5x risk
Protective Factors
- Multi-Product (3+): 0.2x risk
- High Engagement (>70 score): 0.3x risk
- Recent Activity (<7 days): 0.25x risk
- High Balance (>₦500k): 0.4x risk
Data Distributions
Demographics
| Metric | Value |
|---|---|
| Avg Age | 35 years |
| Gender Split | 52% M / 48% F |
| Urban Rate | 65% |
Product Holdings
| Product | Prevalence | Avg Balance |
|---|---|---|
| Savings Account | 80% | ₦150,000 |
| Current Account | 35% | ₦300,000 |
| Debit Card | 40% | - |
| Digital Wallet | 30% | ₦50,000 |
| Loan | 20% | ₦500,000 |
| Investment | 12% | ₦1,000,000 |
| Credit Card | 8% | - |
Average Products per Customer: 2.0
Balance Distribution
- Distribution: Lognormal
- Median: ₦50,000
- Mean: ₦160,000
- Top 10%: >₦800,000
Transaction Activity (30 days)
| Metric | Value |
|---|---|
| Avg Transactions | 25 |
| Zero Transactions | 15% |
| High Activity (50+) | 10% |
Digital Engagement Score
- Distribution: Normal (μ=55, σ=20)
- Low (<30): 20%
- Medium (30-70): 60%
- High (>70): 20%
Customer Segments
| Segment | Prevalence | Avg Balance | Products | Churn Rate |
|---|---|---|---|---|
| Mass Market | 60% | ₦50k | 1.5 | 8% |
| Emerging Affluent | 25% | ₦300k | 2.5 | 4% |
| Affluent | 10% | ₦2M | 3.5 | 2% |
| Private Banking | 5% | ₦10M | 5.0 | 1% |
KYC Tiers
| Tier | Daily Limit | Prevalence | Churn Rate |
|---|---|---|---|
| Tier 1 | ₦300k | 45% | 8% |
| Tier 2 | ₦5M | 45% | 4% |
| Tier 3 | Unlimited | 10% | 2% |
Nigerian Context
Geographic Distribution
Top 5 States:
- Lagos - 25%
- Abuja (FCT) - 12%
- Rivers - 8%
- Kano - 7%
- Oyo - 5%
Income Patterns
- Median Income: ₦120,000/month
- Lagos Multiplier: 1.5x national average
- Rural Areas: 0.6-0.7x national average
Banking Behavior
✅ Salary Detection: 55% have direct deposit
✅ Multi-Banking: 40% hold accounts at 2+ banks
✅ Mobile First: 70% primarily use mobile app
✅ Dormancy: 12.5% dormant (90+ days no activity)
Seasonal Patterns
- Detty December: +50% balances in December
- January Broke: -30% balances in January
- Mid-year Lull: -10% activity in June-July
Files
customer_360/
├── README.md
├── nigerian_customer_360_pilot.parquet (10k, 0.45 MB)
├── nigerian_customer_360.parquet (2M, ~160 MB) - Coming Soon
├── nigerian_customer_360.csv (2M, ~320 MB) - Coming Soon
└── customer_360_sample.csv (100 rows)
Usage Example
import pandas as pd
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.model_selection import train_test_split
# Load data
df = pd.read_parquet('nigerian_customer_360.parquet')
# Feature engineering
df['balance_per_product'] = df['total_balance_ngn'] / df['products_count']
df['txn_per_day'] = df['transaction_count_30d'] / 30
df['avg_txn_size'] = df['transaction_volume_ngn_30d'] / df['transaction_count_30d']
# Churn prediction features
features = [
'products_count', 'total_balance_ngn', 'transaction_count_30d',
'days_since_last_transaction', 'digital_engagement_score',
'dormant_flag', 'account_age_days', 'balance_per_product'
]
X = df[features].fillna(0)
y = df['churn_30d']
# Split and train
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, stratify=y, random_state=42
)
model = GradientBoostingClassifier(n_estimators=100, learning_rate=0.1)
model.fit(X_train, y_train)
# Feature importance
importance_df = pd.DataFrame({
'feature': features,
'importance': model.feature_importances_
}).sort_values('importance', ascending=False)
print(importance_df)
Retention Strategies
High-Risk Segments
- Dormant Customers: Re-engagement campaigns, special offers
- Single-Product: Cross-sell campaigns
- Low Balance: Savings incentives
- Low Engagement: Digital onboarding support
Recommended Actions
- Churn Score >0.7: Immediate intervention (personal call)
- Churn Score 0.5-0.7: Retention offer (fee waiver, bonus)
- Churn Score 0.3-0.5: Engagement campaign (push notifications)
- Churn Score <0.3: Monitor only
Limitations
- Snapshot Data: Single point-in-time, not time-series
- Simplified Engagement: Real engagement includes many more touchpoints
- External Factors: Macroeconomic conditions not included
- Competitive Actions: Other banks' offers not modeled
Citation
@dataset{nigerian_customer_360_2025,
author = {Electric Sheep Africa},
title = {Nigerian Banking Customer 360 Dataset},
year = {2025},
publisher = {Hugging Face}
}
Last Updated: 2025-10-19
Status: ✅ Pilot Validated