Loans and financial matters

Having done what they can to prevent fraud from happening in the first place, banks face the problem of detecting fraud either when it is attempted or after the event to prevent repetitions. In order to do this banks have implemented sophisticated computer systems to identify potentially suspect transactions. These rely on a combination of two factors:
Statistical sampling and analysis.These systems search for unusual transaction behavior, for example a sudden increase in activity on a card that is only used infrequently, or a transaction well above usual levels. Improbable events. These systems look at multiple transactions and attempt to identify whether they are mutually incompatible. It is completely feasible that I could pay a hotel bill in London on Friday and buy an expensive camera in New York the following day. It is infeasible that someone could pay a hotel bill in Hong Kong and simultaneously buy a mobile phone in Italy. (This actually happened to me and the issuing bank caught the infeasible transactions.)
The gap between the large sophisticated issuers and those of small, inexperienced banks new to this business is illustrated with this example. I once visited a bank in Taiwan that was too small to be of interest to institutional investors but was a possible target for a bid. They had fairly recently entered the credit card business as an issuer and were very pleased with the growth in number of cards issued that they had been able to achieve. Their approach to fraud detection was breathtakingly simple. Once an hour they would print out a record of all the transactions over a certain minimum dollar value that had been automatically approved by their systems in the last hour. A handful of individuals then went through these printouts looking for any suspicious transactions. If they found one they then called the cardholder concerned to check whether the transaction was legitimate.