Cross-Sell Loyalty & EBU

Problem & Business Impact

Cross-sell opportunity to existing customers is significantly under-exploited in the telecom sector. Loyal customers and enterprises (Enterprise & Business Unit - EBU) represent the highest-value segments with best receptivity to complementary offers, but untargeted approaches generate low conversion rates (5-8%) and can harm customer relationships. A mobile customer may need fixed Internet, TV, IoT, or additional lines, but these opportunities are rarely identified and activated at the right time.

The EBU segment has specific needs: multi-lines, M2M/IoT, cloud solutions, security. The ability to identify the right decision-maker, the right time (expansion, new site), and the right solution differentiates high-performing cross-sell programs. Our ML approach combines analysis of current service portfolio, expansion signals (usage growth, new sites), and complementary service patterns.

Measured impact:

• 56% increase in cross-sell rate (B2C: 8% → 12.5%, EBU: 15% → 23.5%)
• 24% increase in average value per multi-service customer
• Generation of €7.8M in additional annual revenue
• 18% reduction in churn on multi-service customers

Data & Key Features

Main data sources

• Service portfolio: Active services, tenure, utilization, ARPU per service
• Usage behavior: Consumption patterns, trends, seasonality
• Loyalty data: Tenure, NPS, support interactions, rewards program
• EBU profile: Company size, sector, employee count, multiple sites
• Expansion signals: New sites, hiring, usage increase

Engineered features

• Composite loyalty score (tenure + NPS + engagement)
• Service gaps (typically associated services not subscribed)
• Propensity by service category (fixed, TV, IoT, cloud)
• Optimal timing based on customer lifecycle
• For EBU: Expansion potential score (growth, sector)
• Similarity to multi-service profiles (clustering)
• Bundle and promotion sensitivity
• Predictive customer lifetime value

B2C vs EBU segmentation

• B2C Loyalty (60%): Customers >2 years, NPS >40, stable usage
• B2C High-Value (20%): Top 20% ARPU, high multi-service potential
• EBU SME (15%): SMEs 5-50 employees, growth potential
• EBU Enterprise (5%): Large enterprises, complex needs

Models & Methods

Multi-tier approach by segment

1. Gradient Boosting (Propensity models): Cross-sell probability by service

   • Separate models: HBB, TV, Additional lines, IoT/M2M, Cloud
   • Average AUC-ROC: 0.81 (range: 0.76-0.86)
   • Key features: Current services (31%), loyalty (24%), usage patterns (22%)
   • Top 15% precision: 42-48% by service

2. Matrix Factorization (Collaborative filtering): Similarity-based recommendations

   • "Customers like you also subscribed to..."
   • Detection of optimal bundle patterns
   • Enables identification of non-obvious opportunities

3. Graph Neural Networks (For EBU): Enterprise relationship analysis

   • Mapping of related entities (subsidiaries, sites, decision-makers)
   • Prediction of group-level needs
   • Identification of consolidation opportunities

Cross-sell strategies by segment

• B2C Mobile-only: HBB, TV, convergence proposition
• B2C Fixed-only: Mobile, family multi-lines proposition
• B2C Partial: Bundle completion (mobile+fixed → +TV)
• EBU: Additional lines, IoT/M2M, cloud/security solutions

Real-time Integration

Intelligent recommendation pipeline

Trigger events (anniversary, promo end, new site) → Propensity scoring
    ↓
Optimal service selection → Bundle offer generation
    ↓
Channel routing (email, call center, account manager) → Personalized presentation
    ↓
Conversion tracking → Multi-touch attribution → Model feedback

Optimal intervention moments

• Customer anniversary: Loyalty offer + cross-sell opportunity
• Promotion end: Sustainable bundle proposition at competitive price
• Life events: Relocation, birth, business growth
• Usage signals: Mobile data saturation → HBB proposition
• For EBU: New site, hiring, geographic expansion

Offer personalization

• Intelligent bundles: Complementary services with coherent discount
• Try-before-buy: Trial periods for new services (TV, IoT)
• Smooth migration: Technical facilitation (installation, portability)
• EBU business case: Quantified ROI, sector case studies

KPIs & Performance Targets

| Metric | Target | Current | |--------|--------|---------| | B2C cross-sell rate | >12% | 8.0% → 12.5% | | EBU cross-sell rate | >22% | 15.2% → 23.5% | | Multi-service ARPU | >€65 | €52 → €64 | | Cross-sell revenue | €7M/year | €7.8M | | Multi-service customer churn | <8% | 12.5% → 10.2% | | Model precision (top 15%) | >40% | 44% |

Deployment & Monitoring

Recommendation architecture

1. Feature engineering: Service, usage, loyalty aggregation (daily batch)
2. Multi-model scoring: Propensity by service + ranking
3. Bundle optimization: Optimal combination selection (price, margin, conversion)
4. Intelligent routing: Email (digital natives), Call center (seniors), Account manager (EBU)
5. Attribution tracking: Impact measurement by channel, offer, segment

Performance monitoring

• Dashboard: Opportunities detected, conversions, revenue by service
• Cannibalization analysis (new services vs existing upgrades)
• Post-cross-sell satisfaction tracking (NPS delta)
• A/B testing of bundles and messages
• Monthly retraining with conversion feedback

EBU-specific optimization

• Account manager dashboard with prioritized opportunities
• Alerts on expansion signals (new site detected)
• Playbooks by industry sector (retail, manufacturing, services)
• Wallet share penetration analysis (uncaptured potential)

FAQ & Prerequisites

Q: How to differentiate cross-sell from upsell? A: Cross-sell = selling a service from a different category (mobile → fixed). Upsell = upgrade within the same category (20GB → 50GB plan). Our system handles both but with different models.

Q: How to avoid cannibalizing existing revenue? A: Incrementality analysis: measurement of net additional value. Bundles are calibrated to maximize total revenue, not each service individually. Test/control groups for validation.

Q: What's the approach for complex EBU customers? A: Combination of ML (opportunity identification) + human in the loop (account manager). The system prioritizes and alerts, the manager finalizes with contextual business knowledge.

Q: How to handle customers already multi-service? A: Two approaches: (1) Protection/retention (identify partial unsubscription risks), (2) Expansion (additional unsubscribed services). The model segments according to portfolio maturity.

Q: What are the data prerequisites? A: Unified 360° customer view (all services), detailed usage history, loyalty and NPS data, for EBU: firmographic data (size, sector, structure). CRM integration with opportunity management.