The Hidden Influencer Problem: Using Knowledge Graphs on Snowflake to Uncover Network-Driven Revenue

Most organizations still define customer value through a narrow, transactional lens. Customers who spend the most are classified as VIPs, prioritized for retention programs, and targeted with premium experiences. This model implicitly assumes that revenue is generated independently by individuals, and that each customer’s value is fully captured by their own purchase history.

In reality, this assumption rarely holds. In many industries, such as travel & hospitality, retail, financial services, and subscription platforms, customer behaviour is deeply social. Decisions are influenced by friends, families, and communities. Some customers spend relatively little themselves, yet they play a central role in organizing groups, influencing purchasing decisions, and driving loyalty across entire social circles. If these customers churn, a disproportionate amount of revenue often disappears with them.

This creates what can be described as the Hidden Influencer Problem: the customers who have the greatest long-term impact on revenue are frequently invisible to traditional analytics, because their value is indirect, relational, and structural rather than transactional.

Why Traditional Analytics Is Structurally Insufficient

Relational analytics systems are fundamentally designed to answer questions about direct interactions. They excel at aggregating transactions by joining tables, and producing descriptive metrics such as revenue, frequency, recency and lifetime value. However, they model the world as a collection of records connected by explicit relationships defined in a schema.

This approach breaks down when the underlying problem is not about transactions, but about networks. Concepts such as influence, leadership, community formation and word-of-mouth propagation are not first-class entities in relational systems. They must be approximated using heuristics, thresholds or manually engineered logic, and even then, only direct relationships can be reliably captured.

As a result, most enterprise analytics platforms can measure who bought what, but cannot answer more strategic questions such as:

• Who drives decisions within customer communities?
• How does loyalty propagate through social groups?
• Which customers act as structural anchors for revenue ecosystems?
• What is the real revenue at risk if a socially central customer leaves?

These are not questions of scale or performance. They are questions of data representation. Influence is not a property of rows. It is a property of networks.

Knowledge Graphs to the Rescue

To model influence effectively, customer data needs to be represented as a network of relationships, not just as flat transactional tables. In a knowledge graph, customers are treated as connected entities, allowing the system to explicitly capture how individuals are linked to each other over time. And how strong or weak these bonds between the entities are.

Customer transaction data, behaviour data & demographics data, can all be transformed into this graph structure by identifying repeated co-interactions, shared journeys and consistently common behavioural patterns. Relationships are not created from one-off coincidences or anomalies, but from meaningful signals such as customers traveling together multiple times or exhibiting similar spending behaviour across experiences. This ensures that the model reflects real social connections, rather than accidental overlap.

The resulting structure is no longer just a dataset of purchases and bookings. It becomes a representation of the social fabric of the customer base showing how communities form, how loyalty spreads and how decisions are influenced collectively. This is the critical shift: business can now move from descriptive ‘What’ based questions to diagnostic ‘Why’ based questions and get holistic answers based on collective network strength and not just an individual.

Identifying & Attributing Influence Within Customer Networks

Once customer behaviour is modelled as a network, influence can be measured in a structurally meaningful way. Instead of evaluating customers solely on individual activity, the system evaluates them based on their position within the broader social network.

This makes it possible to distinguish between customers who are merely active and customers who are structurally important. Those who connect many others, shape group behaviour and act as central points of coordination within communities. These customers often sit at the centre of large social clusters and play a critical role in keeping those groups engaged.

This leads to a fundamentally different definition of customer value. A customer becomes valuable not just because of how much they spend, but because of how much collective behaviour can they drive. Their importance comes from the role they play in sustaining and influencing an entire ecosystem, not from their individual transactions alone. These so-called "Influencers” are key for any business to identify and emulate across their larger customer base.

At its core, this connects directly to the fundamental marketing principle of attribution that helps in understanding what truly drives revenue. Is the value attributed solely to the end customer who completes the purchase? To the salesperson who influenced the decision? Or to the often-invisible network influencer whose credibility and relationships ultimately tipped the outcome in your favour?

With the power of the Knowledge Graph, organizations can now answer these questions with clarity and confidence. It not only identifies these hidden influencers but also quantifies their true contribution, enabling a far more accurate and holistic view of revenue attribution and the ecosystem that drives growth.

Quantifying Indirect and Viral Revenue

When customer data is analysed using a relationship-driven, network-based approach, revenue can be measured in a more realistic and strategic way. Instead of attributing value solely based on individual transactions, the business can estimate the economic impact a customer generates through social influence and group behaviour.

Under this approach, customer value extends beyond personal spend and includes the broader revenue associated with their connected network. 

This creates a measurable concept of:

• Total ecosystem value
• Indirect revenue contribution
• Revenue at risk due to churn
• Influence leverage (network value divided by personal spend)

In practice, this reveals customers who may appear low-value in traditional reports, but are in fact critical growth drivers within the ecosystem. These customers act as natural influence channels, sustaining long-term revenue through trust and collective decision making, a dimension of value, that is not captured by conventional transactional analytics.

Graph-Enabled Enterprise Intelligence Platform on Snowflake

Snowflake serves as the centralized foundation for this solution, enabling a seamless transition from transactional systems of record to a governed, business-semantic gold layer that establishes a trusted, enterprise-wide source of truth.

The entire solution, from data preparation to relationship modelling, semantic understanding and AI-driven reasoning, can be implemented natively within the Snowflake ecosystem.

Snowflake provides a single execution environment where enterprise data can be transformed into relationship-aware representations, enriched with semantic context, and analysed through intelligent interfaces. This eliminates the need for external graph databases, specialized AI platforms, or complex data movement pipelines, while ensuring that all intelligence is built directly on governed, production-grade data.

Several Snowflake-native capabilities make this possible. Snowpark enables advanced data processing and feature engineering directly inside Snowflake, allowing complex relationship structures and influence metrics to be derived from raw enterprise data. Snowpark Container Service allows hosting non-native libraries like NetworkX for visualizing Network Graphs on top of the Snowflake hosted enterprise data. Cortex Analyst provides semantic models over both traditional relational features and relationship-driven metrics, enabling natural language interaction with business concepts rather than raw tables. Cortex Search introduces RAG capabilities, enabling semantic search across unstructured data such as documents. Cortex Agents orchestrate reasoning across structured and unstructured layers, allowing autonomous interpretation of business questions and dynamically selecting the most appropriate analytical pathway. Complementing this, Snowflake Intelligence provides a native user interface for business users to seamlessly interact with and reason over their data directly within the Snowflake platform, without the need for additional tools or licensing.

Together, these capabilities position Snowflake not merely as a data platform, but as an end-to-end intelligence system where structured analytics, relationship-aware insights, and AI-driven reasoning coexist natively, on the same data foundation, under a single governance and execution model.

From Data Warehouse to Relationship Intelligence

At Tredence, we help organizations evolve from transaction-centric analytics to relationship-driven intelligence by designing and operationalizing knowledge graph architectures on enterprise data foundations.

Rather than treating customers as isolated records, we enable businesses to model their customer base as an interconnected ecosystem, capturing how value emerges through social influence, shared behaviours, and community dynamics. This allows organizations to move beyond measuring individual transactions and begin understanding how revenue, loyalty, and risk propagate structurally across customer networks.

For our clients, this has unlocked a new class of business capabilities, including:

• Influence-based customer segmentation, identifying structurally important customers beyond traditional spend metrics
• Network-driven retention strategies, prioritizing customers whose departure would create disproportionate downstream revenue impact
• Social ecosystem risk modelling, quantifying revenue exposure embedded within customer communities

From a technology standpoint, this work establishes a modern analytical foundation where:

• Knowledge graphs address strategic questions that cannot be solved using relational analytics alone
• Relationship-derived metrics such as influence, dependency, and ecosystem value become first-class analytical constructs
• AI-driven systems reason over connected entities, enabling more contextual, explainable, and business-aligned intelligence

Through this approach, Tredence enables enterprises to operationalize relationship-aware intelligence directly within their governed data platforms. This represents a fundamental shift from analysing customers as independent transactions to understanding how value is created, sustained, and amplified across real-world networks.

Case Study: How Tredence delivers Knowledge Graph-Driven Root Cause Intelligence for Global Supply Chains

Knowledge graphs sit at the core of how we deliver next-generation operational intelligence. We use them as the primary foundation for end-to-end, AI-driven root cause analysis across complex supply chain and inventory ecosystems.

For a global manufacturing enterprise, Tredence established a comprehensive inventory knowledge graph to unify siloed data across suppliers, SKUs, manufacturing plants, quality systems and distribution networks. This graph explicitly models real-world relationships between operational entities, creating a connected, semantic view of how inventory flows, where dependencies exist, and how issues propagate across the network.

On top of this knowledge graph, we deployed a generative AI-powered RCA layer that reasons over relationships, not just metrics. Business users can query the system in natural language to understand KPIs such as stockouts, excess inventory, Days Inventory Outstanding (DIO) and fill rate, and receive explainable root cause insights grounded in the underlying graph structure.

Some quantified business outcomes translated into:

• 80% reduction in time-to-RCA for inventory anomalies
• Significantly faster identification of systemic drivers behind stockouts, excess inventory, and fill rate degradation
• Measurable improvements in decision velocity and operational responsiveness across global supply chain functions

By combining domain ontologies, relationship modelling, and generative AI, Tredence enables a shift from transactional reporting to knowledge graph-driven operational intelligence, where supply chain performance is continuously understood as a system of interconnected causes and effects, not isolated data points.

AUTHOR - FOLLOW
Arvind Ramachandran
Director, Data Engineering