TSB 2018: the core banking migration that failed

01 · CONTEXT

An entire bank changes platform

Over the weekend of April 20 to 22, 2018, TSB moved about 5 million customers to a new platform, designed, built, and operated by SABIS, the technology services provider of the Sabadell group. The source for this case is the report by the firm Slaughter and May, an independent review commissioned by the TSB board.

The report describes the migration of the functionality and data of an entire bank to a substantially new platform. The deadline also weighed in: Sabadell set the expectation of completing the program by the end of 2017, and the pattern of fixing the date and then building a plan to fit it marked the rest of the program.

1. Before the main migration, mobile app transactions were processed by the old LBG platform.

2. Customer records were migrated to the new platform, and that part went as planned.

3. In the main migration, the app was reconfigured to talk to the new platform's back-end applications through the new middleware.

4. The report's conclusion is that the new platform, as a whole, was not ready to sustain the entire customer base.

02 · THE SINGLE-EVENT CHOICE

Fast and cheap, but with a condition

One of the central decisions in any migration is whether to do it in phases or as a single event, the so-called big bang. The report acknowledges the advantage of the single event and the condition it imposes.

According to the report, TSB did not give sufficient consideration to this choice, which was not substantively discussed by the board. The same report states that the single event would not necessarily have been wrong if the right mitigations had been in place.

03 · THE EARLY TRANSITIONS

Partial proofs before cutover

TSB planned to use the so-called Transition Events to reduce the risk of a migration that would otherwise be entirely a single event. They took two forms: Early Cutovers, which put functionality live earlier, and Live Proving, a pilot with employees.

The Live Proving pilot, in the unscripted phase called TSB Beta, had only about 1,689 participants, all TSB employees. The report assesses that, in many cases, the volumes were too small to prove the functionality and that the participants were unlikely to be representative of the customer base.

For the report, the Early Cutovers did not significantly reduce the risk of the main migration, and that risk would only be mitigated if the platform were subjected to rigorous testing.

04 · THE TESTS

Testing one part and trusting the whole

Functional testing took 17 months, far more than planned. Non-functional testing, which served to confirm that the platform as a whole would operate at the expected service levels, was mostly done in a very compressed period shortly before Go Live.

1. Performance testing was conducted in a single data center.

2. The other data center was left out of the performance test.

3. There were problems in the configuration of the two data centers, which contributed significantly to the digital customers' problems right after Go Live. With testing in only one, it was impossible to identify them beforehand.

4. In the Internet Banking and app tests, the targets were lowered after the tests failed at the original load, and the actual volumes after Go Live exceeded the lowered targets.

The summary of results did not make clear that the targets had changed, and so the results presented a more positive picture than was justified. The board also did not receive an accurate picture of outstanding defects: the actual number was at least two and a half times what was reported to it.

05 · THE IMPACT

What customers saw after Go Live

On the night of Sunday, April 22, 2018, TSB put the new platform live. The migration of customer records went as planned, but many customers had significant problems accessing and using their accounts online or through the app.

On the first day after Go Live, the average wait for a phone agent was about 90 minutes. By May 2, 2018, TSB had received 33,101 complaints, more than 10 times the usual level, and opportunistic fraud attacks reached about 70 times the usual level.

06 · WHAT THE REPORT CONCLUDES

Two causes and insufficient oversight

The report concludes that the problems arose because the new platform was not ready to sustain the entire customer base and because SABIS was not ready to operate it.

It also points out that the TSB board did not seek independent advice on the program as a whole and that TSB did not carry out comprehensive due diligence on SABIS's capability.