Are you a current Kount customer? If so, please select which AWC Environment, Production or Sandbox

If not, please contact our sales department for more information.

Home > Customer Resources > Fraud Management Console > How To Use Our Solution > Boost Technology and the Safety Rating

Boost Technology and the Safety Rating

Overview

Boost Technology aggregates millions of transactions and their outcomes, including approvals, chargebacks, refunds, and reviews. Boost is designed specifically to weigh the risk of fraud against the value of the customer. It helps identify good customers, bad customers, and fraudsters. The machine learning at the core of Boost technology works out-of-the-box, growing smarter as it builds a unique picture of each merchant’s business.

 

The Safety Rating, powered by Boost Technology, is currently in limited availability beta release, with general availability expected during the first quarter of 2018.

What is the Safety Rating?

Transaction safety is the inverse of transaction risk. For example, a safe transaction is not risky and a risky transaction is not safe.

 

The Safety Rating is an indicator of a transaction’s safety relative to all the transactions that pass through Kount. A safe transaction will have a relatively high Safety Rating and an unsafe transaction will have a relatively low Safety Rating. 

 

The Safety Rating ranges from 1 (unsafe) to 999 (very safe). 

 

Mathematically, the Safety Rating represents the percentile of safety for an individual transaction compared to all other transactions that Kount analyzes. For example, a Safety Rating of 100 would indicate the transaction is safer than only 10.0% of all transactions and a Safety Rating of 950 would indicate the transaction is safer than 95.0% of all other transactions.

Statistics

  • Collects 75+ raw data points and converts them into 150+ identifiers
  • From day one, transactions are compared to over 100 million historical records
  • Pushes results through over 250 statistical models
  • Weighs the risk of fraud against the past and future value of the customer
  • Combines with Kount Complete’s entire suite of fraud protection to provide a decision in a fraction of a second

How Does Boost Technology Work?

Via supervised AI Machine Learning, an ensemble classifier with 250 models generates a Safety Rating that indicates the safety of a transaction relative to all the transactions passed through Kount. This is like asking 250 different fraud experts how likely a transaction is to be fraudulent, and then allowing them to vote.

 

The models are designed to identify unique fraud signatures as merchants decision transactions and different fraud patterns emerge. 

Where Does Safety Rating Data Originate?

Response Inquiry Service (RIS) posts and Device Collector data—a set of data that is uniquely available to Kount--are married together, and then normalized into data that can be consumed and analyzed by Kount’s proprietary machine learning algorithms. Boost data is augmented by hundreds of millions of historical transactions generated over a period of six months.

How Does Boost Technology Complement the Persona Risk Score?

Kount’s existing Persona Technology uses one-shot, online learning to detect when fraudsters are using someone’s payment method before anyone else knows it has been stolen. It connects otherwise isolated transactions to give merchants a clear picture of their customers’ online financial Persona. Persona information is collected and delivered in parallel to the data used in Boost Technology.

 

Boost Technology combines transaction details with months of historical customer data to feed an ensemble of statistical models trained on past outcomes. It protects customers, holds bad customers to a higher standard, and finds hidden indicators of fraud without blacklists or whitelists.

 

There are several distinct differences between the Persona Risk Score and Safety Rating:

  • Inverse relationship: Transaction safety is the inverse of transaction risk. The higher the Safety Rating, the safer the transaction; the higher the Persona Risk Score, the riskier the transaction.
  • Assigned Value: The up-to-three-digit Safety Rating correlates to all the other transactions that Kount sees, allowing merchants to assign informed thresholds based on a measure of safety.
  • Data volume: The Safety Rating is ideal in an environment in which the merchant has limited transactional data, because it leverages six months of historical Kount data.

Common Questions

How Often Are Models Trained? 

More than 250 models are trained weekly

 

What specific data is considered?

All high-value data is evaluated, including previously decisioned orders (all automatic decisions, manual reviews, and chargebacks), RIS, Device Collector, Melissa Data (address and geography), and IP address data.

 

What data is not considered? 

Currently, Persona data, cart data, user-defined fields (UDF), third-party callouts, and distance calculation are not considered in the Boost evaluation.

 

How is the Safety Rating Different from the Persona Risk Score?

Data linking doesn’t create the Score. Instead, the Score is generated by past decisions, chargebacks, and other transaction-specific attributes. 

 

Is the speed of the RIS response impacted in any way?

No.

 

How does the Safety Rating complement the Persona Score?

Since the score and the rating consider different data elements, they complement one another in the same way as having two analysts with different criteria weigh in on the same transaction might complement one another. One Score may be higher and one may be lower and together they would indicate the same: higher risk, lower confidence; lower risk, higher confidence.

 

Can I add the Safety Rating as a column in Suspect Orders?

Currently, the Safety Rating is not available in Search, RIS response, or Link Analysis.

 

What Does Boost Cost?

This is a free beta release through the end of the beta period.

 

Who should use the Safety Rating?

Merchants whose Review queues become difficult to manage during high-volume holiday seasons, merchants with excessive chargebacks or false positives, and merchants who rely heavily on VIP black and white lists may benefit from writing rules around the Safety Rating.

 

How do I interpret the Safety Rating for my transactions?

Some merchants find it useful to evaluate Datamart data to arrive at a threshold for decisioning by examining past decisioning patterns, and then selecting a target Score for review and decline rules. The same methodology may apply to the Persona Risk Score.

 

Where do I Locate the Rating?

Boost technology can be leveraged through the rules engine, auto agent rules engine, via Datamart, and from the Transaction Details screen when manually decisioning transactions.

Use Case: Extensive Review Queue

When a review queue becomes unmanageable, either due to high volume holiday seasons or because a rule review is necessary, write auto-approve and auto-decline rules based on historical analysis, the goal of which is to reduce the number of transactions held for manual review.

 

Before writing the rules, a merchant will want to identify a threshold for decisioning, ideally using their past decisioning patterns as a guide.

Use Datamart to Help Identify a Decisioning Threshold

Depending on the merchant’s transaction volume, it is prudent to review two weeks to one month's worth of manually decisioned transactions and based on their Datamart-reported scores, identify a threshold for auto-approve and decline. The same methodology can be applied for analyzing data for both Persona Risk Score and Safety Rating.

 

Note: Historical data is available in Datamart backdated to October 7, 2017.

 

To set a decisioning threshold by analyzing historical decisioning data in Datamart, add the new Boost Safety Rating columns to an analyzer report and analyze as usual. Based on the merchant’s individual scenario, a merchant might elect to write two new rules, automatically declining transactions with a Safety Rating of 100 (for example) or less, and automatically approving transactions with a Safety Rating of 600 (for example) or greater.

Create a No Change Rule to Test Thresholds

Prior to applying thresholds to all transactions, create a No Change rule to test the results without impacting transactions.

 

  • If the merchant determines, based on running the No Change rule, that they will effectively eliminate a percentage of their highest scoring transactions from manual review, they may wish to create a rule to auto-approve those transactions using the Safety Rating threshold. 
  • If the merchant determines, based on running the No Change rule, that they will effectively eliminate a percentage of their lowest scoring transactions from manual review, they may wish to create a rule to auto-decline those transactions using the Safety Rating threshold. 

Use Case: Excessive Chargeback or False Positive Rates

Some merchants might be unsatisfied with their decisioning based on an excessive chargeback rate. If their basic toolset is not sufficiently eliminating false negatives and they are approving fraudulent transactions manually or via rules, incorporating the Safety Rating into their decisioning will improve their chargeback rate.

 

Using Datamart, run a report that displays number of transactions manually reviewed and their respective decisions vis a vis those transactions that were approved and later resulted in chargebacks. Review the Safety Rating and Persona Risk Score of those transactions that were wrongfully approved, and create a rule that automatically flags transactions within a given range for review. 

 

Conversely, if the issue is that of excessive false positives, and merchants are denying quality customers, increase good transactions by creating a rule to automatically approve transactions that were declined that were, in fact, good transactions with sound Safety and Persona Risk Scores.

Use Case: Stringent VIP Lists

There is an inherent danger in creating too stringent a VIP list in that the VIP rules supersede existing rules.  Consider the implications of adding the email address of a trusted client to a VIP Approve List only to learn that client became the victim of identity fraud, in which case all orders from the compromised email address will pass through Kount undetected. Reduce reliability on VIP lists by creating rules around the Safety Rating, instead. 

Administering the Safety Rating

By default, the Rating is calculated for all merchants. Merchants may request visibility into their Safety Ratings by speaking with a Client Success Manager. 

 

You must to post a comment.
Last modified

Tags

Classifications

This page has no classifications.