I’m well known for not liking analogies. I find they generally give people comfort that they understand something without actually changing how much something is understood.
So if I’m forced to use an analogy I’ll at least try to use one that hasn’t been used before, and to use it until it breaks by folding backwards on the analogy so it no longer makes sense. My data quality assurance analogy at the moment is:
Imagine you’re asked to prove that you own your house.
This is an analogous to the regulatory trajectory in financial services – where increasingly data provided to regulators must be attested to met certain data quality criteria.
So again, imagine somebody has asked you to prove that you own your house. You can do this by presenting a deed of title. You might also make a humorous distinction between you owning your house versus you owning a mortgage. Because really the bank owns the house, am I right?
But within this distinction you can make a fairly precise statement about how much of your home you own. You might need to rely on estimates regarding what it’s worth, but you can get the percentages of ownership pretty accurate.
But imagine if deeds of title didn’t exist. Imagine mortgages didn’t exist. Imagine plans that show houses appearing on lots with specific boundaries and reference points for context didn’t exist.
Imagine again being asked to prove that you own your house without the benefit of deeds, mortgages, plans, addresses, and other context. It’s still possible to prove ownership. Now you have to lean on concepts like homesteading; and create a narrative chain of ownership based on the initial claiming and working of the land, through successive transfers of ownership to your own claim. You also have to devise your own way of identifying your house – perhaps using a flag with your family crest.
The problem with this approach to proving ownership is that it’s different for each home. Everybody would need to tell the entire story of how this particular home has come to be on this particular block of land, and who participated at every step of construction and transfer of ownership.
The depth and level of corroboration for this story of ownership would mean we’d need to bring in many of the people who are characters in the narrative and confirm their roles and recollections. Some of these people would disagree with particular points in the story enough to open up doubt or all least require further alternative corroboration.
Once some of the people in the narrative die, or even if they just refuse to turn up for each successive re-telling of the ownership narrative, you lose the ability to prove ownership. This type of approach is therefore clumbersum – requiring a complex narrative that is different for each house – and ultimately inconsistent in the level of assurance it can provide.
The level of assurance is itself dependent on the unique and total narrative around ownership. If, for a particular home, part of the ownership story contains the unsolved murder of the owner and subsequent homesteading by a mysterious stranger, then the certainty of ownership is different than for an ownership story that doesn’t contain that feature. So the idea of a proof with 95% certainty cannot be committed to in general.
The alternative – when you don’t need a completely different narrative ownership story per individual home – can’t be designed by any individual home owner. Instead it has to be built up, shared, agreed, and sustained by the community.
The system for proving home ownership that we have now, that allows for proof of ownership, and even allows as to manage precise percentages of ownership, is the analogy I use for data quality. Because information passes through the community like the ownership of a house, there needs to be a framework agreed by the community so data quality can be consistently understood.
When somebody visits your house for dinner, it is enough that you answer the door to prove sufficient ownership of this house to not expect dinner to be interrupted. Sufficient ownership for this purpose isn’t even real ownership – it could just be a rental agreement. Whereas other assertions of ownership require further proof.
If your organisation doesn’t have artefacts that describe the structure and flow of information it’s like not having house plans that show which property we are talking about. Likewise, if the community doesn’t agree to a specific, potentially costly, process of verification of data as it is transported across the organisation, this is like not having title deeds that you can depend on.
Still with me on this analogy? No, me neither – which is why I don’t like analogies.