Flaminem Case Study

Flaminem Supercharges the Know Your Customer (KYC) Process with Neo4j

Embedded graph technology enables France’s largest public sector investment bank Bpifrance to realize 90% time savings with real-time cascading risk scoring.

Money laundering concerns financial services institutions worldwide. The United Nations estimates that criminal enterprises, terrorists, and corrupt officials launder USD$800 billion to $2 trillion annually, accounting for up to 5% of global GDP.

Flaminem aims to reduce that number. Founded in 2013, the French software vendor is building a new generation of tools to digitize and accelerate processes for Know Your Customer (KYC), fraud detection, anti-money laundering (AML), and Counter-Terrorist Financing (CTF), also known as Combating the Financing of Terrorism (CFT). Flaminem’s tools are crucial for banks looking to stay compliant and secure as financial regulations tighten and money laundering schemes become more complex.

Banks and financial firms are legally required to know their customers’ identities and monitor their financial activity to spot risks. This KYC process can be time-consuming, especially when analyzing large datasets. Financial institutions want to avoid delays while onboarding customers — but failure to conduct due diligence might damage reputations or result in regulatory sanctions.

“We needed a database that would make it simple and fast for our financial service firm clients to evaluate shareholding data and fulfill KYC requirements,” says Antoine Rizk, CEO of Flaminem. Rizk knew that the right technology could give financial firms a critical edge in the balancing act between speed and accuracy. “But we knew we couldn’t track complex shareholder relationships using a relational database due to limitations with that technology.”

Relational databases struggle with nested, interconnected data structures like shareholding networks. A relational database for a shareholding network requires complex joins between tables, leading to poor performance as data complexity grows and more joins are required. In contrast, Neo4j’s graph database doesn’t require joins at all: it natively models relationships, making it easy to traverse ownership structures and quickly adapt to new requirements.

Above: In a Neo4j graph database, each data point, such as an account holder name or an address, is called a node. Neo4j graphs natively model relationships between nodes, making it easy to traverse financial ownership structures.

Neo4j Powers Flaminem’s KYC, Enabling Real-Time Risk Scoring

Flaminem designed its prototype KYC platform using an open-source Apache TinkerPop framework that could be plugged into any graph database. Performance concerns however soon led Flaminem to focus its efforts on Neo4j:

“We found that an organizational tree structure does not work for shareholding analysis. There are so many intersecting webs of relationships that graph provides the best approach,” says Stéphane Debart, CTO, Flaminem. “With that in mind, we selected Neo4j for our next proof of concept.”

Flaminem upgraded from Neo4j’s Community Edition to its Enterprise Edition to access additional features needed for its enterprise application, such as:

• Clustering to provide fault tolerance and automatic load balancing for high availability. Flaminem’s tools are business critical for the customers they serve — so in the rare event of hardware or network failures, users can still access their data and use the tools without disruption. The cluster’s causal consistency, also known as session consistency, also guarantees that the database write operation that created a user account, for example, is present when that same user subsequently attempts to log in.
• Online backup and recovery operations to avoid database downtime, enabling Flaminem to better support stringent SLAs for its customers. 
• Advanced Cypher capabilities such as node and relationship key constraints, which serve as guardrails that help Flaminem maintain data integrity. 

Cypher is Neo4j’s declarative, schema-flexible graph query language. This means that users are not required to use a fixed schema to represent data and that they can add new attributes and relationships as their graphs evolve, as well as adapt to changing regulations across countries and global financial systems.

“As a software provider, we value Neo4j’s schema flexibility, which makes our solutions easily configurable for varying client data models,” Debart says.

Flaminem also values the ability to add client-specific labels in each Neo4j deployment. Flaminem’s customers can use their own business terminology and product-oriented labels, or apply labels in different languages like French or English, with the power to perform Cypher queries across vocabulary sets. 

“This customizability sets up smoother client communications. No matter the complexity, we can typically load graphs in a single Cypher query. Our clients grasp things faster because we embrace their vocabulary,” says Debart. “Our clients love that shareholder graphs are easy to understand. Neo4j accelerates development, enablement, and adoption. It also helps us as we look to adapt our tools for use with other clients and involve consulting partners.”

Bpifrance Achieves 90% Time Savings with Flaminem’s Solution

Flaminem today manages 50 Neo4j graph databases for more than 20 customers on its OVHcloud environment. Its solutions are invaluable to clients like Bpifrance, France’s largest public-sector investment bank.

Bpifrance’s collaboration with Flaminem began with a small KYC proof of concept. Within just one year, Bpifrance had rolled out Flaminem’s solution to all departments worldwide. More than 3,000 analysts and business users at Bpifrance now use the KYC platform daily to calculate shareholder risk scores. Flaminem cascades this risk scoring across its entire Neo4j database in less than a second. If a customer becomes negatively exposed and creates greater risk, their score is re-calculated in near-real time.

“Bpifrance found that the tool empowers business users and analysts. The graph makes a very profound productivity boost possible,” says Rizk. “The cascading risk calculations produce a 90% time savings for the bank’s analysts.” 

Before Flaminem KYC, only specialized analysts could process due diligence requests. The result created bottlenecks that frustrated customers. Bpifrance account managers and generalists can now use Flaminem’s KYC solution to onboard customers, reducing the time it takes for the bank to conduct due diligence and freeing analyst time for more complex fraud detection. “Bpifrance improved its compliance while providing faster customer service and reducing costs,” Rizk notes.
 
The success of Flaminem’s KYC tool paved the way for new products and solutions. In 2024, the company introduced additional support for fraud detection and case navigation, integrating visualization capabilities from Neo4j Bloom. These new features allow Bpifrance analysts to explore complex data relationships visually, making it easier to identify fraud rings and anomalies in credit requests. The graph-based UI has proven effective in helping analysts quickly grasp the context behind high-risk alerts and pinpoint fraud.

“We are excited to improve our services with more graph-oriented visualizations. It makes our client solutions that much more compelling and valuable. It’s a real differentiator to have Neo4j at the heart of our platform,” says Rizk. “Our clients say it’s a primary reason they choose to work with Flaminem. Our relationship with Neo4j benefits our business and our clients at the highest level.”

Get in Touch

Curious about what insights you could unlock for your business with graph-powered solutions? Let’s talk — reach out, and we’ll get in touch.