Multi-Portfolio Acquisition

Problem & Business Impact

Untargeted customer acquisition generates high costs and low conversion rates. Operators spend massively on marketing without optimizing targeting by segment and portfolio.

Measured impact:

• +15-20% conversion rate on targeted campaigns
• -20% customer acquisition cost (CAC)
• +25% quality score (high LTV customers)
• Campaign ROI: 5:1 vs. 2:1 in generic approach

Data & Key Features

Main data sources

• Demographics: age, socio-professional category, location, estimated income
• Browsing behavior: site/app visits, pages viewed, duration
• Interaction history: call center, retail, emails, SMS
• Geolocation data: residential area, frequented places
• External signals: DMP, social networks, partnerships

Engineered features

• Propensity score by portfolio (prepaid, postpaid, enterprise)
• Recommended product (optimal plan)
• Preferred acquisition channel (web, retail, phone)
• Price sensitivity (premium vs. economy)
• Optimal contact timing

Models & Methods

Multi-level approach

1. Gradient Boosting (XGBoost): Multi-class classification

   • Prepaid vs. Postpaid vs. Enterprise
   • Accuracy: 84%
   • Main features: demographics, behavior

2. Neural Networks: Optimal product prediction

   • Architecture: 3 hidden layers
   • Recommendation accuracy: 79%
   • Complex signal integration

3. Propensity Scoring: Conversion probability

   • Score 0-100 by segment
   • Regular calibration
   • Dynamic thresholds per campaign

Intelligent segmentation

• Young urbanites: data-centric prepaid, mobile app
• Families: postpaid multi-lines, bundles
• Professionals: business plans, VAS services
• Seniors: simplicity, retail proximity
• Enterprises: enterprise solutions, account management

Real-time Integration

Multi-channel orchestration

Scoring engine
     ↓
Lead prioritization (top 20% propensity)
     ↓
Intelligent routing:
├── Web: Display ads, retargeting
├── App: Personalized push notifications
├── Retail: Prospect list for advisors
├── Call center: Personalized scripts
└── SMS/Email: One-click offers

Dynamic personalization

• Segment-adapted offer (price, data, voice)
• Optimal channel by preference
• Timing (avoid undesirable hours)
• Personalized creative (language, visuals)

KPIs & Performance Targets

| Metric | Target | Current | |--------|--------|---------| | Conversion rate | +15-20% | +18% | | CAC reduction | -20% | -22% | | Quality score | +25% | +27% | | Model accuracy | >80% | 84% | | Time to convert | <14d | 11d |

Deployment & Monitoring

Acquisition pipeline

1. Daily scoring: Prospect base refresh
2. Segmentation: Segment + optimal product assignment
3. Prioritization: Top 20% propensity score
4. Activation: Coordinated multi-channel campaigns
5. Tracking: Conversion, attribution, D+90 LTV

Systematic A/B Testing

• Tested variables: Offer, channel, timing, creative
• Minimum duration: 2 weeks
• Metrics: Conversion rate, CAC, LTV, D+90 churn
• Winner deployment: Automatic if uplift >10%

Feedback loop & improvement

• Campaign results integration in retraining
• Propensity threshold adjustment based on performance
• Post-acquisition analysis (churn, ARPU, satisfaction)

FAQ & Prerequisites

Q: Minimum data required? A: Prospect base (min. demographics + contact), 3-month interaction history.

Q: Compatible with existing CRM systems? A: Yes, REST API + batch export for Salesforce, Siebel, custom integration.

Q: GDPR / opt-in? A: Scoring on marketing opt-in base. Pseudonymization. Right to object respected.

Q: Implementation timeline? A: 2-3 months (data pipeline + CRM integration + team training).

Q: Expected ROI? A: 5:1 over 12 months (vs. 2:1 untargeted campaigns).