Fraud Detection: How to Detect Fraud in a Company

Increased detection means finding more fraud sooner. Rarely do we see organizations that closely track fraud losses on a daily basis (and make changes to their detection systems as a result), but also know how to adjust their efforts as they see things going awry.

In the extreme, if we were to remove the detection tools and functions from an organization, we would have to wait for customers to tell us that their credit has all been used up or that their statement had arrived with transactions that they do not recognize. These can then be investigated, refunded, charged-back, etc., but the losses are going to be far too high. Often in other industries there are more challenging issues of detection - i.e. insurers often NEVER see the loss as it gets paid in and amongst the big claims payouts that are made.

In contrast, if we saw a customer with daily use of their card for six consecutive days in an ATM in Canada for the maximum amount, we might investigate this by calling the customer — or just block the card when we see that it has also been used in Sainsbury's in the UK in between these ATM transactions. An even better scenario would be when the first ATM transaction occurred only three hours after the last UK card-present transaction — in which instance the fraud could seemingly be stopped before we lose any money. Similarly, goods sent out en-masse from a retailer from internet orders should cause some conce
as should repeat address changes, or repeat refunds or insurance claims claims.

So this in turn leads us to the transaction scoring systems, clever and clear analysis of patterns, rules built into our authorization systems, earlier action upon compromises, tracking of transactions and looking for patterns. The solutions can get quite complex, as they should be.

If card businesses rely upon only one system, or one process based upon 'quarterly updates' in scorecards, then they will not be doing enough. Today we should be looking at fraud that happened yesterday in order to make detection tools better tomorrow, and in real time where possible.