How Embedded Payments uses Validation Services and Machine Learning to combat fraud

Businesses want to offer their clients a powerful way to control payouts, and J.P. Morgan Embedded Payments helps optimize money movement and payment timing. However, fraudsters increasingly target marketplaces, contractor platforms, and bill pay systems. That's why organizations on B2B platforms and third-party marketplaces will want to mitigate fraud risk. One solution: Our validation services, which integrate into our embedded payments. Read on for how this protection can help you combat fraud and cyber challenges during onboarding, account linking, and ongoing payment flows. 

Validation Services: The backbone of scalable payments

J.P. Morgan’s Validation Services are API-driven tools that integrate directly into your platform’s onboarding and payment flows.  

There are two primary types of validations performed:  

• Account Validation (our focus today): Confirms the legitimacy of financial accounts using multiple sources like account and transaction data. Methods include direct data checks, microdeposits and name matching. 
• Entity Validation: Verifies individuals or businesses using details such as name, address, government ID and cross-references with 3rd party providers and our own screening. 

This image highlights validation services in action: 

Technical Deep-Dive: Name matching during Account Validation

Let’s deep dive into how Embedded Payments leverages Account Validation Services with name matching for account linking – critical when sellers (C2s) link external bank accounts for payouts.

There are 3 major steps taken to name match efficiently and intelligently:

1. Name cleanup: We use a clustered engine to standardize and clean account names correcting spelling errors, word order, and truncations. This helps ensure consistency and accuracy before validation.
2. Name classification: To process external account linking efficiently, we tag each account name as either individual or corporate. Our name classification engine uses custom text features and an LSTM model (a deep learning technique) to accurately classify names and support streamlined operations.
3. Name matching: The cleaned and classified name is then validated against the name on record, helping to verify that the account truly belongs to the intended user. This is where our word-level similarity matrix and the Munkres algorithm are applied to conduct one-to-one verification checks. 
   • How it works - score calculation: Scoring to reflect quality of matches
     1. Generates an assignment list that pairs each word in the provided name to a unique word in the profile name, maximizing overall similarity.
     2. The score calculation then evaluates these matched pairs:
        • If there is not a perfect match the algorithm leverages the Munkres assignment and calculates the final score using their similarity values.
     3. This helps ensure that only the best, non-overlapping matches are considered, and the scoring reflects the quality of these matches.

This process happens in near real-time during account linking, providing feedback to platforms and their clients. As a result, you reduce manual intervention and mitigate risk of fraudulent account linking.

Business Outcomes: Why it matters

• Fraud Prevention: Machine-learning driven validation can block suspicious accounts before they are linked.
• Operational Efficiency: The name matching process minimizes manual reviews and errors, accelerating payouts and improving the user experience.
• Enhanced Trust: Strong controls boost user confidence, driving platform adoption and retention.

Embedded Payments: Developer-ready, enterprise grade

Partner with J.P. Morgan Embedded Payments for a strategic advantage. Whether onboarding sellers, paying suppliers, or funding borrowers, our tools help you move money with confidence. 

What’s next? 

Stay tuned for follow-up blog posts exploring additional Account Validation features that are leveraged in Embedded Payments.  

Ready to build with us? Connect with our Embedded Payments team here to learn more.