South Africa’s Standard Bank is expanding its international franchise, and in order to support its international efforts, the bank has reengineered its reference data operations. A key project undertaken over the past year has been to put together a global Client Information File (CIF) in order to provide a single, global view of its clients across the wholesale bank.
Steve Spark, who is running the client information project at Standard Bank, spoke of the aim and progress of the project as well as the issues they faced along the way, at the recent Azdex event.
The project included defining the scope of the CIF system, as well as putting in place the processes and procedures to ensure that the data quality could be pro-actively maintained. Fortunately, the group have a very supportive chief executive. Spark said, “Senior management are now asking about core client data, which is a significant change over last year.”
The first decision made was that a central system to feed target systems in its Johannesburg base as well as London and Hong Kong, was needed. It was also decided that the CIF must be kept as a thin application, which currently supports approximately 30 key core client data fields to support client identification. Said Spark, “The aim of the system is to uniquely identify clients. We do receive internal requests for more data to be stored in the system, which we evaluate but if it does not help to uniquely identify clients then we have to turn them down.”
Some representatives from larger or more global institutions said that such a thin approach could not work for their operations given their wider instrument and market segment coverage.
Undertaken next was a cleansing of Standard Bank’s data files in order to create the single list of global clients. Said Spark, “We had a lot of duplicates to cleanse as our international records would often overlap with our local South African records”. In this endeavour, Standard Bank worked with Azdex to de-duplicate its records. They now total 27,500 records globally held in the central database.
Standard Bank soon recognized the need for regional focus to cater for country-specific data. Spark said, “Partly due to cost and resource issues, but also because of language difficulties in regions such as Hong Kong, we realised that we couldn’t do all the data cleansing and quality analysis from the central location of Johannesburg.”
The team is now working on legal hierarchies, but here, “we continue to struggle,” said Spark. Particularly in South Africa, he says, where there is a lot of acquisition activity and so hierarchies can be difficult.
To support ongoing maintenance of the core client data, a Central Data Group (CDG) with members in each location, has been created. The group is now in the process of “implementing a robust mechanism to publish data to its downstream systems”. These systems – there are currently six on board, the rest will be added over the next two years – will ‘subscribe’ to various data fields of relevance to their processes.
They have put in place daily reporting to measure how many records have changed, and they ensure they adhere to standards and have automated checks as much as possible. There is also a paper trail to see what analysts have changed, information that the bank is considering imaging to put into an online archive.
Other measures such as hard-coding the cities, countries and regions help to maintain data integrity. The processes and procedures are continually updated in Standard Bank’s CIF ‘bible’ of documentation.
There have been compromises along the way, however, in order to cater for operational practicalities, said Spark. “CIF is not completely thin as we have added some non-core data, such as the KYC status, client roles, and system mapping table”. The system mapping table maps all the systems that a client is subscribed to.
In terms of the lessons learned from their initiative, Spark emphasised that “people are the single biggest success factor,” an element that was initially underestimated. “Our mistake was not to realise that this job required more than an inexperienced person just capturing data from the web.” Now the bank looks for people with enquiring minds that will delve deeper into changes and ensure the data is fully verified.
It has also changed its approach from looking for volume in data cleansing, which it realised was the wrong incentive. “Its quality not quantity that matters. We estimated record cleansing would take between 30-40 minutes, but the truth is closer to 60-90 minutes per record, a lot of which is taken up by researching the hierarchy.” He also estimates that it costs around £15 per record to cleanse, not an insignificant amount. And “data changes faster than you would imagine”.
The team structure is also important, where there should be a clear distinction and timing between the roles of cleansing, verification, and authorisation. And as many reference data experts will agree, technology is just a minor component of the effort. Said Spark, “processes and procedures are key and the change management effort is significant”. He warned that anyone embarking on such a project “Will underestimate the data cleansing effort!”