What is customer outcomes data?
At its most simple level, a customer outcome is the result of knowing if a query to Ekata was associated with a “good” or “bad” event related to fraud that the customer might face. For example, if an ecommerce customer queries Ekata on transactional orders within their fraud decisioning workflow, they may later learn that some transactions were fraudulent and resulted in chargebacks while others did not. Linking these transactional level fraud related learnings (i.e., outcomes data) to their original Ekata queries is central to Ekata’s ability to provide highly predictive machine learning features.
Still, as outcomes are defined in many different ways, it is important that Ekata understands how the customer outcomes are determined, collected, and, therefore, defined.
As previously mentioned for ecommerce use cases, “bad” events may be defined by fraud-related chargebacks. For account opening use cases, where customers query Ekata on account sign up, “bad” events may be defined by promotional or rewards abuse, chargebacks, spam, abusive content behavior, and other fraud-related account activity events.
As you can imagine, it can take time before these outcomes are known. A fraudster might sign up for an account but not commit fraud on the platform right away. This period between an event and the final outcome is called the “maturity period,” and its length may vary between different types of outcomes. Chargebacks, for example, have a typical maturation period between 60-90 days. Ekata requests that only outcomes that have aged past their maturation period be shared to ensure the quality of the labels.
How does Ekata use outcomes data?
Ekata uses customer outcome data to increase the predictiveness of our machine learning products. Ekata does not use outcome data to label any person, business, phone, address, email, or IP as potentially fraudulent or create any type of approval list. We solely use the data to improve our ability to identify patterns and retrain and refine our models. Furthermore, all outcome data is hashed and encrypted. Raw outcome data will never be disclosed to any of our other customers and cannot be accessed outside of Ekata.
By providing outcome data on a recurring basis, customers greatly improve the efficacy of Ekata’s machine learning models over time. It’s a unique opportunity for customers to directly improve the products that they use. Our machine learning models improve by learning from the fraud patterns customers are actually seeing and as a result become more refined and predictive with each model retraining. Our machine learning based attributes, such as Identity Risk Score and Network Score, thus provide more lift and value to our customer base. These scores also tend to be the strongest Ekata signals that customers rely on to help them identify both “good” and “bad” actors.
How can current customers share outcomes with Ekata?
Ekata offers a programmatic, secure way for customers to share outcomes data regularly with us via our Outcomes API. After a simple integration, customers can easily send us the outcomes associated to past Ekata queries so they are available for our Data Science team to use to retrain our models.
Current customers who are interested in sharing their outcomes data with Ekata should get in touch with their Account Manager to learn more.