Problem & Business Impact

Acquiring new customers via portability (port-in) represents a strategic growth opportunity, particularly to capture value customers already familiar with mobile services. Unlike acquiring entirely new customers, port-ins generally have a superior value profile (average ARPU +35%) and better initial retention. However, traditional acquisition campaigns suffer from low conversion rates (2-4%) and high CAC.

The challenge is to identify, among a large population of competitor customers, those who are truly in an active consideration phase for changing operators. Our approach combines behavioral signal analysis (web searches, visits, engagement), competitive intelligence (network weakness zones), and propensity scoring models to target prospects with highest conversion probability.

Measured impact:

58% increase in port-in campaign conversion rate
32% reduction in CAC (€146 → €99)
Acquisition of 12,500 new port-in customers/year (+40%)
420% ROI on ML-targeted campaigns

Data & Key Features

Main data sources

Web behavioral data: Google/Bing searches, site visits, comparators
Advertising data: Display/search ad engagement, retargeting
Competitive intelligence: Competitor network weakness zones, public incidents
Demographic data: Age, location, general mobile behavior
Third-party data: Lookalike audiences, intent segments (Google, Meta)

Engineered features

Purchase intention score (NLP analysis of searches)
Interaction frequency and recency (website, ads)
Similarity to successful port-in customer profile (lookalike)
Exposure to competitor network issues (geographic zone)
Profile alignment with our positioning (premium, value, data-centric)
Estimated promotion sensitivity
Conversion probability by channel (digital, phone, store)
Projected lifetime value score

Models & Methods

Multi-stage approach for optimal targeting

Propensity Model (Gradient Boosting): Port-in probability within 30 days
- AUC-ROC: 0.78
- Top 20% precision: 42% (vs 4% baseline)
- Key features: Site engagement (32%), competitor searches (24%), network weakness zone (18%)
- Segmentation: High intent vs medium vs low
Lookalike Modeling (Collaborative Filtering): Audience expansion
- Identification of profiles similar to best recent port-ins
- Use of Meta/Google data for audience expansion
- Enables campaign scaling while maintaining quality
Response Prediction (Neural Networks): Offer and channel optimization
- Response rate prediction by offer type and channel
- Message and timing personalization
- Advertising budget optimization (bid optimization)

Port-in target segmentation

Quality dissatisfied: Competitor customers with recurring network issues (38%)
Promotion hunters: Sensitive to aggressive offers, regular switching (28%)
Upgraders: Seeking better quality/services (22%)
Relocators: Geographic zone change (12%)

Real-time Integration

Targeting and activation pipeline

Web/app signals (real-time) → Propensity scoring (streaming)
    ↓
Profile enrichment → Intent/value segmentation
    ↓
Audience generation (Google/Meta) → Campaign activation
    ↓
Conversion tracking → Model feedback

Multi-channel activation

Digital ads: Display/search retargeting for high-intent segments
Social media: Meta/TikTok campaigns for lookalike audiences
Email/SMS: Direct targeting for known prospects (opt-in base)
Stores: Advisor guidance for identified walk-ins

Offer personalization

High-value prospects: Premium offers, enhanced portability bonus
Price-sensitive: Aggressive promotions, long-term discounts
Quality seekers: 5G network, coverage, services highlight
Data users: Unlimited data plans, international roaming

KPIs & Performance Targets

| Metric | Target | Current | |--------|--------|---------| | Campaign conversion rate | >5% | 2.4% → 6.3% | | Average CAC | <€100 | €146 → €99 | | Monthly port-ins | >1000 | 730 → 1,040 | | Model precision (top 20%) | >40% | 42% | | Port-in ARPU vs baseline | >+30% | +35% | | Campaign ROI | >350% | 420% |

Deployment & Monitoring

Production pipeline

Signal collection: Web analytics, ad platforms, CRM data aggregation
Daily scoring: Propensity calculation for identified prospects
Segmentation: Classification by intent, value, optimal channel
Audience generation: Export to Google Ads, Meta, programmatic DSP
Conversion tracking: Multi-touch attribution, ROI analysis by segment
Retraining: Monthly with actual conversion feedback

Continuous optimization

A/B testing of creatives and messages by segment
Automatic bid optimization based on propensity score
Budget reallocation to top-performing segments
Cannibalization analysis (port-in vs natural acquisition)

Privacy & compliance

GDPR compliance: use of anonymized and aggregated data
Lookalike audiences without PII sharing
Opt-out mechanisms for contacted prospects
Transparency in third-party data usage

FAQ & Prerequisites

Q: How to identify prospects before they become customers? A: Via public behavioral signals: web searches (Google Analytics, search ads), visits to our site/app, ad engagement. Complemented by lookalike audiences built on aggregated profiles.

Q: How to estimate the value of a non-customer prospect? A: Predictive LTV model based on: demographic profile, web behavior, similarity to existing customers, geographic zone. Accuracy: ~65% correlation with actual post-acquisition ARPU.

Q: What's the difference from classic acquisition? A: Port-in specifically targets active customers of other operators (so already mobile users), vs acquiring any prospect. Superior conversion rate (+60%) but more restricted audience.

Q: How to measure ROI while respecting GDPR? A: Aggregated attribution via conversion APIs (Google, Meta), without individual tracking. Cohort analysis and test/control groups to measure incremental impact of ML-guided campaigns.

Q: What are the technical prerequisites? A: Integration with Google Ads, Meta Ads Manager, web analytics (GA4), CRM with lead management, customer data platform for data unification, propensity and lookalike models in production.

MNP Port-In Opportunity