First Look - Kount, by James Taylor, Everything Decision Management


I got a chance to catch up with Kountrecently, a 4 year old company that is a wholly owned subsidiary ofKeynetics(a 13 year old e-commerce company) focused on preventing fraud in e-commerce. They offer an all-in-one SaaS solution aimed at detecting fraud fast (intra-transaction real-time) while giving customers flexibility and usability. They have strategic partnerships with Chase Paymentech (who handle 50% of card not present transactions worldwide and white label Kount for their fraud solution), LexisNexis (both to pull data from their data sources but also as a white labeler of Kount) and Planet Payment (focused on Asia Pacific).

Card-not-present retail fraud cost businesses over $100B in 2010 with losses including unrecoverable products and chargebacks. Fraud, of course, is an arms race between those trying to prevent fraud and the organized crime rings that are trying to make money out of fraud. Each new fraud technique results in a new technology or approach that is designed to address it. This means that most merchants are using 3-5 different fraud tools and still doing a lot of manual review in response to reports that are often after the fact. There’s little strategic approach to this in many merchants with most taking a scattergun approach.

To deal with fraud, Kount identifies a number of fraud screening key technologieslike geolocation, device identification, affiliate management, identify authentication, fraud scores, proxy detection, shared data and identity verification. These need to be pulled together with a strategy for 3rd party data, with a set of merchant-specific rules and black lists/white lists layered on top and integrated with operational management of reviews/cases etc. Their solution address three pieces – technology, strategy and operational management

Kounts’ basic technologyruns a bunch of analytics for device finger printing, geolocation, dynamic scoring, cross-merchant linking etc. These elements are provided to their decision engine directly and combined into a risk score – the Kount score. Kount ensures this score is consistent so that an 86 means more or less the same over time and across users for instance. These analytics, merchant data, variable data and the risk score are passed to the Kount decision engine. This generates a Yes/No/Review decision and does so in 250-350ms. This gives a very fast yes/no for most transactions.

The Review decision however prompts their automated agent that calls out for additional data and makes another attempt to auto adjudicate or to finally refer to a manual decision. This takes another 1-3 seconds and essentially provides an automated but slower back up to increase the rate of auto decisioning without slowing down the obvious ones. This is a nice feature as all too often the first automated decision in a system bounces straight to a manual review and a second, slower but more sophisticated check is a useful addition. The two-layer approach allows them to hit 95% in real real-time with another 3-4% coming from the auto agent in 1-3 seconds.

They also emphasize the fact that they integrate the technology so it is acoherent set. Combined with the SaaS model and a simple interface to call the service it is an easy to set up environment and easy to integrate with the customer’s own agents. Interestingly they continue to monitor transactions for several weeks to help spot “personas” or indirect links between cards and transactions across merchants. And this means that even if the original transaction seemed ok the system may come back later as subsequent transactions raise the likelihood of that initial transaction being fraudulent.

Customers see both top and bottom line impact – less fraud, more completed transactions, lower costs, faster processing time, better bottom line. For example a telco customer started with over 1% chargeback rate that meant they were out of compliance with the credit card fraud programs so they were paying $100,000/month in fines from Visa. 30 days after partnering with Kount they were off the Visa list and by 60 days had reduced fraud by 75%. They are currently running at 0.07% chargeback with manual review of only 1.2% transactions having eliminated 90% of their fraud. In fact Kount is so confident they offer a 60 day guarantee!

Kount provides a web-based case management system for agents to use to manage reviews and see re-scored transactions etc. though companies with their own case management system can integrate with the fraud detection system to manage cases in one place. More sophisticated companies can integrate their order processing system, for instance, using an XML interface.

Rules are managed through a web interface too with companies able to add their own approve/decline/review rules with their own reason codes and messages. These rules can access the analytic attributes built by Kount – scores, locations, velocities across multiple cards or devices (what they call a persona) etc. They may also be part of rules-based scorecards to create a custom score (combining various elements to create a new score) that is then the basis for further rules. A specific ruleset can be defined for particular days or periods (such as black Friday or Christmas).

Kount is a classic “Decisions as a Service” offering from my perspective – pass in a single transaction, ask the question “how should I handle this transaction” and get back an Accept/Reject/Refer decision. It provides some further integration as well as a case management system but you could, and many do, use it purely as a decisioning service.