ShopperDNA

Many risk mitigation platforms work under the assumption that a transaction is a standalone entity. More advanced systems cluster transactions based on the user's login credentials or credit card number. Fraudsters use a multitude of techniques to circumvent this kind of clusters, usually by using new compromised credit card numbers and changing devices and login identities. 

ShopperDNA builds on this idea of clustering using advanced linking algorithms to track fraudsters even as they change devices, networks, and identities. In this, the ShopperDNA clusters a merchant's transactions by the ShopperDNA shopper that committed those transactions. This shopper entity is dynamically re-evaluated and can change as the ShopperDNA system receives more information about the user's behavior. The power of this system is that it allows the creation of automated rules that monitor the behavior of this shopper, effectively eliminating the ability of fraudsters to hide their behavior across different transactions. 

How ShopperDNA works

The goal of the ShopperDNA algorithm is to identify and link these transactions that, most likely, are generated by the same shopper. Both fraudsters and trusted users.

 

The figure above displays a network of transactions generated by the same shopper. Transactions are shown as squares and the linkable attributes (also called identifiers) are shown as circles. Each linkable attribute (such as an IP address) is continually assessed for its uniqueness and this affects the weight it gives to any given transaction connection. 

The ShopperDNA Algorithm Logic is as: 

  1. All transactions that share attributes with the parent transaction are determined.
  2. The strength of those attributes is tracked (algorithmically based on various uniqueness and temporal signals).
  3. All transactions that mean a dynamic confidence threshold are linked and considered part of the same shopper entity.

Things to keep in mind

  • Seemingly high confidence attributes may not be used for linking in some scenarios. For example, when an IP address has been identified as being a shared address or a credit card has been identified as being shared, like in an office setting. 
  • A ShopperDNA shopper is always a malleable construct. ShopperDNA networks change over time and should be considered simply another way to track users, in addition to other methods like shopper reference or name.