PiCARD^TM: IMPROVING THE BOTTOM LINE
A common question heard in business today is, "How can we improve the bottom line?"  It doesn't take a degree in accounting to realize that the bottom line is improved by increasing revenue and/or decreasing expenses. 

Most organizations attempt to reduce expenses first, because they have greater control over cash outflows than they do over cash inflows.  One very popular way to reduce expenses is to focus on strategic cost management throughout the supply chain.  Organizations begin by reducing inefficiencies in their own processes and then concentrate on removing inefficiencies in both their suppliers' and customers' processes.  It is clear that inefficiencies in the supply chain have a multiplicative effect.

THE PURCHASE CARD SOLUTION
Implementation of a purchase card program lends itself to the reduction of expenses. Purchasing professionals spend less time on paper-pushing associated with a  purchase order system, and more time on building relationships within the supply chain.   

Allowing employees the autonomy to make smaller everyday purchases without the purchasing department's direct approval also reduces the amount of time required for administrative tasks. 

These time reductions are gained at the expense of a tradeoff.  Given the loss of central control in purchasing smaller items, there is a greater potential for misuse in a purchase card environment than in the traditional purchase order setting.  PiCard^TM uses AFFM^TM technology to detect and prevents this misuse. 

PROBLEM DOMAIN 
Given that procurement cards are a good way to improve the bottom line, a logical question is then, "What's the quickest, easiest, and least expensive way to deter misuse of the procurement card system?" 

The ideal method of deterrence is to audit each transaction, but clearly this is not quick, easy or inexpensive.  Commonly companies employ random auditing to  combat fraud.  Random audits are easy, quick, and relatively inexpensive and certainly the fear of an audit can deter the potential fraud,  but is the random audit likely to detect actual fraudulent use?  Clearly, a random audit though easy, quick, and inexpensive is no more likely than chance to turn up misuse.
 

A better way to distribute the internal auditors is to audit only novel and rare transactions.  This begs the question, how can a company quickly, easily, and inexpensively target the auditing process toward novel or rare transactions? 

Answer - Automate the targeting using Adaptive Pattern Recognition.
 
WHY AUTOMATE?
The commercial credit industry has established that their customers establish habitual patterns of use and that any substantial deviation from this pattern indicates the possibility of a stolen card or card information. 

Procurement card users will logically develop the same habitual patterns, and  because the procurement card system captures every transaction electronically, it  makes sense that an automated fraud detection method would be feasible. 

Automating the targeting process saves time by reducing the number of necessary audits.  This reduced number of audits will maintain the same level of deterrence and take less time. Furthermore, if fraud is present, the targeted audits are likely to detect more fraud then random audits.

WHY ADAPTIVE PATTERN RECOGNITION?
The main problem with traditional (supervised) fraud detection systems is that they rely on historical examples of both fraudulent and non-fraudulent use.  They also require that the ratio of fraudulent to non-fraudulent transaction be around 1 to 1, when in practice the number of fraudulent transactions in a purchasing system will significantly less than the number of normal transactions.  AFFM^TM technology is designed to detect rare and novel events and does not operate under the same conditions as supervised modeling  methods. 

The second major shortcoming of supervised modeling methods is that they are static.  Once trained, they are fixed and only able to accurately recognize behaviors presented to them during development.  This will cause behaviors not represented in the training population to be misclassified. AFFM^TM technology merely categorizes new transactions as belonging to an existing group, or as rare or novel.  This focus on rare and novel transactions allows it to target "suspicious" transactions in real time. 

A third problem with traditional techniques is that the models they produce are difficult to explain in real world terms.  AFFM^TM technology flags rare transactions and allows an auditor to quickly determine why the transaction was not like the others.  Additionally, auditing directors will have the ability to determine what the largest contributors or indicators for misuse are: valuable insights for directed auditing.

ADDITIONAL INFORMATION
To initiate a discussion regarding the potential role of PiCard in your environment, please extend a call invitation, or send an Email request.

* PiCard and AFFM are trademarks of American Heuristics Corporation