Back to basics big time

The big US fines around data capture illustrate an interesting dichotomy: while banks and technology companies work on sophisticated analytics, AI, and behavioural models, the basics of data capture are still unfinished. There are technology providers who believe that instead of trying to pipe existing data feeds into clever new aggregation and analytics boxes, banks must go back to the beginning and replace outdated systems to build the datasets needed for automated trade reconstruction, integrated surveillance and for delivering true, new business intelligence. Garbage into the smartest AI system will still be garbage out. And there are banks on that three-to-five-year journey. But not many.

Those datasets are also at the heart of assurance. So, should Boards move away from thinking about better compliance and towards ensuring that they know their risks and have certainty over the results coming out of their systems? One panellist said, “The Board should really be focusing from an assurance perspective and we should be kicking the tyres on it internally. I would far rather be building and showing the regulators at a much slower pace, a way of sort of capturing the stuff in the old-fashioned, more manual way, than trust that the data is right rather than switching on a machine and expecting it to find stuff from around the world through open source and you are going to trust the data.”

The big question is this: given the cost and complexity of getting data right, who will do it without being forced to? Some major institutions have been driven back to basics but most have not. And this is not just a big bank problem. Smaller institutions sometimes have even worse legacy systems and data problems than their more complex cousins. Regulators seem content right now to pile new obligations onto the 3 lines of defence without prescribing better data capture, management and governance. At what point do they and the industry accept that the compliance edifice built over the past 10 years implies a fundamental rethink of data and everything that rests on it?