Following the merger of Credit Union Australia (CUA) and the Australian National Credit Union (ANCU) on January 1, the process of integrating 400,000 customer records is being slowed by manual data mapping.
The merger will see the creation of the country's "biggest credit union network", known as CUA, with 75 branches, 800 staff, and some $5 billion in assets under management.
Barry Los, CIO of the combined entity, said the ANCU merger is akin to a "convergence project" where data is being migrated from the Fiserv financial application to Abacus on an IBM mainframe.
"It really isn't a software matter, but one of extracting data and uploading it to the Abacus databases," Los said, adding the project is manually intensive because the integration team has to do a data mapping exercise.
It's this manual integration that is resulting in services being the most significant expense for the project, not software.
As to whether improved support for open data formats by software vendors will help alleviate such integration woes in the future, Los was sceptical.
"The chances of [common data formats] happening are remote, because the databases are from two different sources anyway," he said.
Hydrasight research director, John Brand, said large integration projects associated with mergers and acquisitions are a CIO's "worst nightmare", because of the sheer complexity of the data sets.
"There is a misconception that a customer is a customer is a customer," Brand said. "When you get complex relationships between customers, and the contracts they have, vendors struggle with integrating those."
Brand said vendors give a "good impression" of integration, but it is just an initial step that, once overcome, gives users a path to integrate the data.
"What constitutes a customer is more complex than a name and address; whatever data environment you are migrating to must fit the exceptions to what products can be offered to what people," he said. "It's not as simple as just migrate people to one system."
Brand agreed with Los' claim that services, not software, will make up the most significant proportion of the integration cost, because modelling "doesn't do anything, but shows how to do it" and therefore the services markets are "taking up this challenge".
"It's never the structure of the data that's the problem, it's the semantic meaning," Brand said. "Technology vendors don't know what it means, [because] they provide the data models and say fill in the field of customer information."
According to Brand, there is also an assumption of one system going into another system during a merger, "but that's never the case".
"Integration and transformation doesn't happen once as there are many layers of transformation that has to occur [and] you have all these complex data dependencies," he said. "Architecture tools don't make the process of discovering how data is used any easier."