Not All False Positives Are Bad
0 comment(s)
Here at Memento we have a thing about false positives. Like all of you, we just don't like them. But before we go too far down that particular rant we need to step back and realize that not all false positives are bad. Keep in mind that a false positive occurs when there is a logical explanation for something that looks odd, or shouldn't have happened. Often it's just that the fraud detection rules need adjustment or improvement; but sometimes there are causes that present an opportunity for operational improvements. Deep down, I've known this for a long time from my experiences with fraud detection, but a recent visit to a customer pointed it out to me in crystal clear fashion. We were looking at a list of alerts and the analytical technician was not happy with the false positive rate. So, we did some root cause analysis.
Duplicate serial numbers:
- A read error caused the serial number to be read as one that had already posted. This can happen from a damaged MICR line, poor check quality, dirty read head, or I suppose a number of other technical causes I am not aware of. Expect this once in awhile, but if you see a lot of them, let somebody know so they can take corrective action.
- This particular serial number was 9999. The most common cause is a telemarketer creating a remotely created check and not getting a serial number from the customer, which by the way is counter to the regulation. If you see this a lot from one particular source, let them know. If it's your customer counsel them. If it's from a particular bank of first deposit, ask them to counsel their customer. Remotely created checks are notorious for this kind of thing, and are a significant source for fraudulent items as well.
- It turns out that in this case the customer actually generated 2 valid checks with the same serial number, one from a checkbook and one from a PC. This customer needs to be counseled about safe banking practices.
This one turned out to be an encoding error. The presenting bank is probably sitting on a credit difference in their adjustments department waiting for you to ask for it (we all work debits first). Get with the exceptions department to make the correction and prevent improper returns of that or other checks, and get it corrected before it impacts the customer.
A misread account number (MICR line damage from fold) caused a misposted item (reconcilement wouldn't have caught this one). Again, get with exceptions to correct this situation before there is customer impact.
Other fixable causes of false positive rates might include:
- A manually returned item that was not included in the return system's feed to the fraud system will cause a duplicate serial number when the presenting bank re-presents the item because the history was not updated to reflect the return. Correct the procedure.
- Serial out of range being caused by the bill pay system generating checks that they assign their own serial number to (this can also create duplicate check numbers). Get with your bill pay folks to either convert those transactions to ACH or tell you what checks they generated (like an internal positive pay)
- The use of multiple checkbooks that overlap within the history retention period will cause duplicate serial numbers. Customer needs to be educated about the proper selection of serial ranges for multiple checkbook use. If you don't talk to them you will see these every day.
- Counter checks can cause "no serial number" or "duplicate serial number" false positives depending on how you configure your counter checks. This is a thorny problem. The best answer is, don't use counter checks! But if you must, have severe restrictions on their use. These false positives, as bad as they are, are a tool for monitoring adherence to the policy.
As you can see, some false positives give you the opportunity to avoid a bank error that impact customers, correct a bad procedure, or take other corrective action for an operational customer service problem. We still don't like the unhelpful false positives... those are still bad.