In a classic TV exchange, less than one day into the new government administration in what seems like an age ago now, White House spokesperson Kellyanne Conway asserts “the president was giving his alternative facts”. NBC News anchor Chuck Todd, clearly baffled angrily asserts that such a thing does not exist.
It all makes for entertaining TV. But how often do we come across “alternative facts” in the real world outside cable TV news shows?
Back to reality with a bump – consider a security descriptive attribute ‘Is Financial’. This flag indicates whether the security held is in a financial company. AIFMD, Form PF filings require fund managers to make this distinction. In the real world the underlying definition may vary depending on who is asking. A fund manager’s definition hinges around her investment mandate, or possibly the compliance department’s memos. The regulatory team may have a different definition. The regulators themselves, across different jurisdictions, may vary yet again. Clearly, even in something as simple as a Yes/No flag, there is more than one version of the truth.
The perceived wisdom says the way to solve this problem is a technique that has been around for decades: mastering. Master Data Management is the process of constructing a central master copy of an organization’s data. In this paradigm, our Yes/No flag would be replicated for each consumer: Is Financial (Fund Manager), Is Financial (Compliance), Is Financial (Form PF) and so on. But is this sensible? How many versions of the same flag will we end up needing?
This problem amplifies when more complicated data elements are considered. Consider position nominal value. The back office team uses specialist software to calculate this at least every night and rolls this information up to the fund level. This software – a fund accounting or portfolio management system – is highly specialized. Changing it is risky and expensive. Should we really be overloading it with multiple values when definitions vary depending on which end-consumer it ultimately serves?
Master Data Management is a concept that looks great on paper but rapidly falls apart upon contact with reality. We think the reason for this is confusion between what systems of record need to store and what different consumers need delivered. Too many data management systems are obsessed with storing and not enough about delivering. Modern data management techniques let us define semantics at the last mile as it is delivered to the system that is asking for it. After all, in the real world, there is more than one version of the truth – it all depends who is asking.