In Data Governance in Context, Jim Ericson outlines several paths of data governance, or as I put it: Who Do You Say You Are?:
On one path, more enterprises are dead serious about creating and using data they can trust and verify. It’s a simple equation. Data that isn’t properly owned and operated can’t be used for regulatory work, won’t be trusted to make significant business decisions and will never have the value organizations keep wanting to ascribe it on the balance sheet. We now know instinctively that with correct and thorough information, we can jump on opportunities, unite our understanding and steer the business better than before.
On a similar path, we embrace tested data in the marketplace (see Experian, D&B, etc.) that is trusted for a use case even if it does not conform to internal standards. Nothing wrong with that either.
And on yet another path (and areas between) it’s exploration and discovery of data that might engage huge general samples of data with imprecise value.
It’s clear that we cannot and won’t have the same governance standards for all the different data now facing an enterprise.
For starters, crowd sourced and third party data bring a new dimension, because “fitness for purpose” is by definition a relative term. You don’t need or want the same standard for how many thousands or millions of visitors used a website feature or clicked on a bundle in the way you maintain your customer or financial info.
Do mortgage-backed securities fall into the “…huge general samples of data with imprecise value?” I ask because I don’t work in the financial industry. Or do they not practice data governance, except to generate numbers for the auditors?
I mention this because I suspect that subject identity governance would be equally useful for topic map authoring.
For some topic maps, say on drug trials, need to have a high degree of reliability and auditability. As well as precise identification (even if double-blind) of the test subjects.
Or there may be different tests for subject identity, some of which appear to be less precise than others.
For example, merging all the topics entered by a particular operator in a day to look for patterns that may indicate they are not following data entry protocols. (It is hard to be as random as real data.)
As with most issues, there isn’t any hard and fast rule that works for all cases. You do need to document the rules you are following and for how long. It will help you test old rules and to formulate new ones. (“Document” meaning to write down. The vagaries of memory are insufficient.)