"Tony, I really appreciate the feedback. As David Hood notes in his post, "The Community Against Fraud”, there is "strong sense of community that exists in fraud prevention”. Sharing your thoughts took time and effort. Thanks for strengthening the fraud community! I think you touched on the Holy Grail when you mentioned allowing "the fraud investigator to interact with disparate data and uncover non-obvious relationships” The 9/11 commission cited the failure of imagination as one of the reasons that the attacks were not thwarted. That may be very true. In more concrete terms, it may also have been the failure to share data between agencies using a technology platform that could gather the data, (disparate or otherwise) and provide the analyst with the beginnings of the picture, or in this case, the indicators of a terrorist attack planned for US soil. However, I think that is important to note that once presented with a "new picture”, investigators may be unable to appropriately interpret the data. With new and improved data analysis tools must come additional training to reorient, or reeducate the investigator on how best to interpret the data on their screen. You noted the following, "I believe that the optimal analytical system includes a combination of rules to detect fraud, automated models to predict fraud and interactive analytics to investigate suspicious activity.” If I understand your comment correctly; the "combination of rules” would be defined by the bank that owns the software. Banking, more than any other industry (and certainly the airlines for regulatory reasons), is more apt to share information about how to detect fraud (See David Hood's article). That's actually where I believe banks have a distinct advantage. Most other industries that I been exposed to are almost completely in the dark about how their competitors detect and prevent fraud. Further, you mentioned "Automated models to predict fraud” presumably that would be tried and tested fraud rules that repeatedly detect fraud? And finally, I completely agree that an optimal system should include interactive analytics. So many fraud detection platforms are remarkably rigid. They do not allow the bank, or user to update the "rules” or data feeds to reflect the latest intelligence. In my opinion, in order to be truly effective, a fraud detection platform should be flexible enough that new streams of data can added with minimal effort, and incorporate rules that can be quickly changed to reflect changes in fraud schemes. Fraudsters do not perpetrate fraud in the same manner month after month, why should banks not demand the ability to change their fraud rules accordingly? I also agree with your statement regarding collaboration. Unfortunately, collaboration within corporate fraud departments can be somewhat lacking. Often there are just not enough hours in the day to allow investigators to weigh in on a fraud investigation being conducted by another member of the team. Better data aggregation, or case management, coupled with the ability to "slice and dice” the data quickly can help leverage the collective knowledge of the fraud department rather than just one, or possibly two investigators allocated to the case. Better yet, allowing investigators to see patterns in different investigations can be an invaluable training tool. On a side note, having met with various federal, state and local law enforcement agencies I can attest to the importance of presenting complex cases in a readily understandable fashion (i.e. visual). There are some extremely smart people in law enforcement (and obviously the intelligence agencies you mention) however, leveraging their knowledge and engaging their assistance is so much easier when you present the case in manner they can understand and then "sell” to their boss, or the U.S. Attorney's office etc. Law enforcement must understand the fraud before they invest their time and effort to pursue. There is no quicker way to lose law enforcement support than to waste their time with an overly complex or badly organized case. Again, thank you for taking the time to comment. You very eloquently expanded on my original post. I truly appreciate the insight, as well as the time and effort you put forth to respond. I look forward to seeing more of your comments! "

"Thank you Paul. I enjoyed your article. From an analytics perspective, it seems to me that there are two major classes of analysis that allow the fraud investigator to look below the water line. Predictive Analytics can be used to build models that predict outcomes such as how much money is at risk or the probability of fraud taking place. This can be effective and is used by some banks today. These models require robust input data sets, period recalibration and expertise to manage the building and deployment of models. You allude to the 2nd analytic technique in your article which we should also discuss. You write: "Use the elements of the accidental discovery to mine the data. Gain the upper hand by adopting an aggressive, proactive approach to detecting transactions with similar elements. The vast majority of frauds have common elements. Use the data to find those elements in other transactions. The discovery of the fraud is on some level, a gift. If the fraud involved a general ledger, mine the ledger for similar transactions." One way to do this is through "Interactive Analytics" or "IA”. IA allows the fraud investigator to interact with disparate data and uncover non-obvious relationships that may exist below the water line. This form of analytics is consistent with the expertise of the fraud investigator and also happens to be how the CIA, FBI and NSA solve mission critical applications in the areas of financial crimes analysis, cyber breaches and homeland defense. In this context, a more appropriate way of describing this technique is "Investigative Analytics". In either case, the core components seem to be 1) fraud analysts visualizing rich picture of alerts, accounts, loan officers, historical transactions and other data in a variety of visual forms to detect non-obvious relationships that may exist below the water line, 2) unifying these data sources inside these visualizations so that a complete picture can be drawn and 3) collaborating with other team members, local law enforcement or government agencies to solve share insights. In some cases, the data unification piece should be done "on demand” when needed. In other words, as the analyst uncovers something suspicious, additional data can be brought into the investigation and incorporated into a single visualization. This often reveals the hidden insights necessary to solve the case. I believe that the optimal analytical system includes a combination of rules to detect fraud, automated models to predict fraud and interactive analytics to investigate suspicious activity. With the appropriate feedback loops to improve rule sets and create new input variables for the predictive models, interactive analytics provides a complimentary approach to these technologies. Paul, thank you for your article. Your approach and thinking is right on target, Fraud continues to become a bigger and bigger problem worldwide. I only hope that others share in this approach and begin to deploy these techniques so that the honest citizens don't have to pay for losses that continue to pile up as a result of this problem. Tony Agresta Centrifuge Systems www.centrifugesystems.com 571-830-1390 "