A software failure and subsequent human oversight were behind the major disruption in OCBC Bank's services for about three hours last Saturday.
Group chief executive Samuel Tsien said yesterday that there was a problem in "the backup unit for the core banking system" that meant customer transaction data was not transferred.
The glitch, which has prompted a response from the Monetary Authority of Singapore (MAS), caused the core banking system's storage to reach maximum capacity. This in turn stopped it from accepting new data, he added.
Mr Tsien said: "The software failure signal was unfortunately not detected for rectification due to a human oversight."
The problem affected OCBC's ATM network, online banking channels, Nets and card services from about 8.45pm last Saturday. They were gradually restored after 11.30pm.
Mr Tsien added in an e-mail statement: "As we restarted the system to restore the services, we made sure that all customer and transaction data was protected."
He said the service disruption was "totally unacceptable" and that measures had been put in place to prevent a recurrence. He apologised to customers for the inconvenience caused by the disruption.
The outage drew complaints from users on social media sites, with customers citing problems from withdrawing cash to getting online access to their accounts.
DATA PROTECTED WHEN RESTARTING SYSTEM
The software failure signal was unfortunately not detected for rectification due to a human oversight. As we restarted the system to restore the services, we made sure that all customer and transaction data was protected.
GROUP CHIEF EXECUTIVE SAMUEL TSIEN
A spokesman for the MAS said OCBC notified it last Saturday that "it was encountering intermittent disruption to its provision of Internet banking, ATM, Fast and card payment acceptance services to customers".
"We have instructed OCBC to identify and address the root causes of the disruption, and submit a thorough investigation report," the spokesman added.
It also said it will closely monitor OCBC's rectification measures following the outage, and where necessary, take appropriate supervisory actions.
The MAS said critical IT systems are required to be tested regularly to ensure that they can resume operations within four hours following any disruption, and to have a maximum unscheduled downtime of no more than four hours across a 12-month period.