Posts Tagged ‘SotirisLyras’
[DevoxxGR2026] An Online Real-World Unified Data Architecture: Bridging Digital Analytics and Transactions for Comprehensive Customer Understanding
Lecturers
Sotiris Lyras and Andreas Athanasopoulos are experienced data professionals working within the online analytics and data science teams at OPAP, a leading betting and gaming operator in Greece. Their work centers on creating holistic views of customer behavior by intelligently integrating digital engagement metrics with transactional and financial records in complex, high-volume, real-time environments.
Abstract
In this practical, production-focused case study presented at Devoxx Greece 2026, Sotiris Lyras and Andreas Athanasopoulos detail the implementation of a unified data architecture that successfully merges digital analytics signals with transactional business records. Beginning with comprehensive instrumentation of web and mobile applications through Google Tag Manager and flowing into BigQuery for transformation and standardization, behavioral data is then joined with financial and account activity within Azure Databricks. This creates a true 360-degree customer profile that powers enhanced segmentation, propensity modeling, churn prediction, customer lifetime value calculations, and personalized experiences—even for unregistered users exhibiting strong digital intent but lacking transactional history.
The Persistent Challenge of Fragmented Customer Data
Modern online platforms typically maintain parallel but disconnected data ecosystems. Digital analytics platforms capture rich behavioral signals including session patterns, device characteristics, navigation flows, and engagement metrics. Meanwhile, core transactional systems record financial movements, betting activity, deposits, withdrawals, and account status changes. Without systematic integration, organizations lack a complete, unified understanding of their customers’ journeys.
This fragmentation manifests in several critical ways. Marketing teams struggle to connect campaign exposure with actual revenue outcomes. Product development operates with incomplete behavioral context when prioritizing features or measuring success. Machine learning models trained on partial datasets deliver suboptimal performance and limited predictive power. Most significantly, users who demonstrate clear digital interest and engagement but have not yet converted to registered, paying customers remain largely invisible to business systems optimized for transactional signals.
Architecting and Implementing the Unified Data Foundation
The solution begins with meticulous, structured instrumentation across all customer touchpoints. Google Tag Manager provides a robust framework for consistent event tracking, ensuring that every meaningful user interaction—whether on web or mobile applications—is captured as an atomic event enriched with relevant contextual metadata.
These raw digital events are ingested into Google BigQuery, where comprehensive transformation, cleansing, standardization, and enrichment processes occur. The team invested significant effort in identity resolution strategies that accurately link user activity across devices and sessions while respecting privacy constraints and avoiding duplication.
Azure Databricks serves as the central integration and unification layer. Here, transformed behavioral data streams are intelligently joined with transactional and financial records at the individual customer level. The architecture maintains rigorous data governance, auditability, privacy compliance, and quality controls throughout the entire pipeline.
Realizing Business Value Through Holistic Customer Intelligence
The unified dataset fundamentally transforms organizational capabilities:
- Complete Journey Analysis: Teams can now analyze how factors such as device type, acquisition channel, personalization segments, and behavioral patterns influence conversion probability, engagement depth, and long-term value.
- Accurate Acquisition Attribution: Digital marketing campaigns can be evaluated based on their actual contribution to first deposits, betting activity, and customer lifetime value rather than surface-level engagement metrics.
- Proactive Unregistered User Engagement: Behavioral scoring models identify high-intent anonymous visitors, enabling targeted nurturing and conversion optimization before formal registration occurs.
- Significantly Enhanced Machine Learning: Predictive models for deposit propensity, churn risk, reactivation likelihood, and customer lifetime value achieve substantially higher performance when trained on rich, combined feature sets.
Measurable outcomes include more precise customer segmentation, dramatically improved personalization relevance, earlier and more accurate identification of at-risk customers, and overall more effective resource allocation across marketing, product, and customer experience initiatives.
Broader Implications for Product Development, Marketing, and Operations
Access to unified customer data accelerates product development cycles by providing complete information for hypothesis validation and success measurement. Customer support and retention teams gain richer contextual understanding, enabling more personalized and effective interventions. Marketing organizations can optimize budget allocation by clearly connecting digital behaviors with downstream business results.
Beyond specific functional improvements, the unified architecture represents a profound shift in organizational perspective. Rather than viewing users through disconnected lenses of “digital visitors” versus “paying customers,” the business now understands each individual as progressing through a single, continuous journey with multiple observable dimensions.
Technical and Organizational Lessons Learned
Successful implementation demanded careful attention to several critical areas: achieving reliable identity resolution across fragmented touchpoints, aligning disparate data schemas and freshness requirements, maintaining strict governance and compliance standards, and fostering close collaboration between analytics, data engineering, product, and business stakeholders.
The speakers emphasize that effective data unification initiatives should begin with clearly articulated business questions and use cases rather than purely technical elegance. They also stress the importance of treating the unified dataset as a strategic organizational asset requiring ongoing investment in quality, governance, and accessibility.
Conclusion: Unified Data as Foundation for Modern Customer-Centric Organizations
A thoughtfully designed unified data architecture converts previously fragmented signals into powerful, actionable intelligence. For organizations operating in data-intensive industries, successfully bridging digital behavioral data with transactional records provides the essential foundation for genuinely customer-centric decision-making across product development, marketing, operations, and strategic planning.
The OPAP implementation demonstrates that even in complex, regulated, high-volume environments, technical data unification can deliver substantial, measurable business impact when approached with clear objectives, robust engineering practices, and strong cross-functional collaboration.