Coris is modernizing risk management for software platforms worldwide. To achieve this mission, we’re constantly building products that address payment fraud risks in a systematic way.

Today, we’re excited to introduce the world’s first credit card transaction model built specifically for software platforms’ unique needs. This novel model analyzes both sides of a transaction – merchant and payer – to identify its potential for a chargeback. 

Credit card fraud is a serious problem, with 63% of U.S. credit card holders stating they’ve experienced fraud at least once. Our model addresses this problem head-on through holistic transaction analysis from the merchant and payer perspectives. This surfaces “bad” transactions more accurately and at a fraction of the time it takes other models, reducing downstream losses and allowing risk teams to operate more efficiently.

Read on to learn more, and contact us to get started today.

‍

Existing credit card models miss merchant-level signals

Software companies often leverage generic payment fraud tools to monitor for credit card fraud on their platform. These models do a good job analyzing payer-level information and highlighting instances of consumer fraud. However, they don’t address the other side of the transaction: the merchant.

Why is this important? Increasingly, platforms are noticing transactions between legitimate payers and suspicious merchants. For example, a software platform that serves salons and spas might notice legitimate consumers transacting with a spa business that has an increasing number of negative Google reviews. This creates financial and reputational risks for the software platform. If consumers are shortchanged by the spa, they will dispute the transaction, and the platform will have to absorb this loss and incur associated chargeback fees.

Currently, software platforms have a hard time systematically identifying this kind of chargeback scenario. Payment fraud tools have existed for years, but they are opaque and typically focused on chargebacks stemming from payer activity. Software platforms lack control over these models and can’t customize them to address their unique transactional needs.

As software platforms increasingly complained about this pain point, we asked ourselves: could we extend our payment risk management and fraud capabilities to the credit card use case?

‍

Automated transaction scoring to reduce chargebacks & false positives

Our credit card model is the first tool to holistically evaluate a transaction for chargeback potential. It integrates with Stripe and Adyen to instantly analyze incoming transactions in-realtime. The model screens each transaction for indicators of chargeback risk, including:

• Payer-level information: payer name, email, phone number, mailing address, and IP address
• Credit card information: The model goes beyond just typical metadata like issuer bank details, AVS match, CVV match, and charge amount. It checks a software platform’s previous transactional history to understand if it has interacted with this payer and credit card before, and surfaces any positive or negative signals. 
• Merchant-level information: The model is the first to incorporate merchant metadata in transaction analysis. It leverages MerchantProfiler to analyze merchant-specific information such as business age, website attributes, online reviews and trends, industry classification, and more. In addition, it checks the purpose of the transaction, whether the merchant has conducted this kind of transaction before, whether the merchant and payer have transacted before, and surfaces any useful historical trends. 

After conducting this analysis, the model assigns each transaction a score between 0 and 100, with 0 indicating low chargeback risk and 100 indicating high chargeback risk.

Risk teams can use these scores to set up custom risk rules and actions in our rule engine. For example, they can automatically decline transactions with a high risk of fraud (where the score is closer to 100), or forward these cases to manual review. They can also combine transaction scores with other signals (such as the payment amount) into one risk rule.

“Fraud prevention for payments on a platform using Stripe Connect used to feel like guesswork—Stripe Radar helped, but it didn’t go far enough. This new fraud model is on another level. It dives into sub-merchant data, evaluates merchant quality, and gives us a complete view of risk across our payments ecosystem. Unlike Stripe Radar, it’s tailored to the complexities of platforms like ours, catching fraud more precisely while reducing false positives. It’s the solution we’ve been waiting for to protect payments and keep our merchants growing safely.”

• Zack Sullivan, Head of Finance at Hearth

As previously mentioned, the credit card transaction model is currently available to customers of Stripe Connect and Adyen for Platforms. Stripe customers leverage the model in two ways: scoring the transaction before capture (allowing the platform to return it before transaction completion), or after capture (refund after the transaction is completed). Adyen customers can score transactions after capture.

‍

Why is this better than existing payment models? 

Incorporating merchant-level information into credit card transaction models is superior for two reasons:

• Merchant history in one place: Software platforms can review all transactions tied to an individual merchant at once, instead of reviewing transactions in a one-off fashion. This allows risk teams to review a merchant’s transaction history to see if they have successfully (or unsuccessfully) processed the same kind of transaction in the past. This allows for more accurate decision-making.
• Fewer false positives: Generic payment fraud models are not customized to the software platform use case, and thus might flag legitimate transactions as suspicious purely based on payer signals. Our model can more accurately contextualize transactions using underlying transaction metadata such as invoice details. These additional signals help reduce the number of false positives forwarded to manual review.

‍

Want to learn more?

Contact us if you’d like to learn more about our credit card model, or if you have a chargeback use case you’d like us to tackle.