Tipalti Takes Down Fraudsters With New Detect Module

Merchants

Tipalti has caught a few bad guys lately — 7,400 of them, to be exact.

The San Mateo startup recently announced the addition of a new fraud monitoring module, Tipalti Detect, to help eliminate accounts payable fraud risk. Tipalti Detect stopped those 7,400 payments throughout its six-month beta phase and first week live, eliminating an estimated $4 million in fraud losses for Tipalti’s customers, according to reports from Rob Israch, the startup’s CMO, during his October check-in with PYMNTS.

Tipalti Detect integrates with the rest of Tipalti’s payments automation solution to track all payee activity and perform in-depth risk checks. It does this early in the relationship rather than waiting until fraudulent payees show their true colors.

Israch explained the module looks for patterns across the supplier database, analyzing factors such as tax data, contact information and IP address, among other elements. If, for example, a payee’s name doesn’t match his or her payment details or tax ID number, Tipalti Detect will flag it.

Of course, because the system is just looking for patterns, Israch said the Detect numbers are much higher than the 7,400 instances that turned out to be fraud. Tipalti Detect is just the first line of defense.

It is up to the payer’s finance team to investigate why the transaction was flagged and whether it wishes to allow or prevent it, Israch explained. The only exceptions are high-risk Office of Foreign Assets Control (OFAC) hits. In these cases, Tipalti will stop the payment from going through because allowing it could mean allowing money laundering or drug trafficking, he said.

While the process resembles what many industry players are doing with machine learning, Israch is careful to make the distinction that Detect is not an artificial intelligence (AI) solution. It is an algorithm, he explained, and can only find patterns that have been fed to it by humans.

However, that’s clearly been enough to put the kibosh on a whole lot of attempted fraud in a very short beta phase and lifespan.