I attended and spoke this week at the International Data Quality Summit organised by the IAIDQ. I had a great time listening to my peers and talking shop over beers and good food. Learned a lot and I hope some people learned something from me.
Now, besides my regular talk I also sat in the IAIDQ Data Quality Panel. And we embarked on something that still resonates a bit with me. The question of the moderator began innocent ‘Define Data Quality’. There were like two schools of quality that seemed to be combatting each other a bit.
On the one hand there was - let me call it - the ISO school of quality.
ISO 9000 deals with quality in general: Quality is the degree to which a set of inherent characteristics fulfils requirements. Where requirement is defined as ‘a need or expectation that is stated, generally implied or obligatory’.
ISO 8000 is somewhat derived from ISO 9000 but more rephrased towards Data Quality: Quality data is data that meets stated requirements.
Now, I had just been to the session of Allan Duncan about gathering information requirements and he said some wise words; be polemic if you want discussion, tension, feedback, learning, etc.. Well, being polemic is something that comes quite natural to me.. So there I go…
The moderator asked basically if I agreed with this ISO definition of quality. Me - being polemic - said “No, I do not agree”. I cited the definition of quality as I learned it from Jerry Weinberg: Quality is value to some person. And I added two rules to this definition that were inspired by a blog of Markus Gartner I read some time ago: the Relative Rule and the Time Rule.
The Relative Rule is obvious; the word ‘quality’ is like ‘complexity’ or ‘beauty’….is has only value in the eye of the beholder. A human being - a person - a warm body….
The Time Rule is somewhat more sophisticated: the value that is being put on beauty, complexity or quality depends on time as well, not only on a human being. Something that is complex now, can be perceived as less complex a year later. Same goes for beauty, something that is beautiful now, can be perceived ugly as hell a year later ;-). And yes, same goes for quality. Something that is perceived as high quality by someone now, can be perceived as being of low quality a year later. So quality in my opinion is never in a fixed state, it is always moving, fluently through time.
So, a bit rephrased; quality is value to some person at some time
There you have it; the ISO-school of quality and - let me call it - the Weinberg-school of quality. I think the ISO-guy was getting a bit irritated with me - not sure though. Be hey - blame Alan - I needed to be polemic to get the discussion going to a place where we all actually learned something.
In the ISO-school opinion you can always state the quality of data by knowing their requirements. In my world (the Weinberg-school) you can never state the quality of data unless you know the person and subsequently what he or she values at that time. Now, to be very annoying. What if we ship a product (data product or any other product) to thousands of people...or even more. Oh crap, to actually know the quality, I now need to know all these people and what they value to have an idea about the degree of quality? Yes, that it exactly my point.
The ISO-school seems to think we can state the requirements finitely and thus can put a value to it. In my world I do not believe that. Or, in other words, I can meet stated requirements (voiced by person X at time Y) but still deliver a crappy quality. Why? Because, we humans are extremely bad at expressing our requirements, validating them and assessing whether or not we are complete in all our requirements. The probability by the way that someone can actually finitely state their requirements is something that I reject. I reject that as a world-view even…..
Final remarks; assessing quality is a moving target. So lets throw in the towel? Nah. Remember, I was being polemic (!)
As Danette McGilvray very eloquently put it in the discussion at the end: on various dimensions of data quality we have a duty to be as precise as we can regarding stating the requirements and subsequently designing systems that can uphold these requirements. Yes Yes and Yes…So Agree.
In data quality we gotta meet known requirements as best we can. But to say that quality is high if we have met these stated requirements? My answer; I do not know. Only person X knows….or Y...or Z...at a certain moment in time.
This blog has inpsired Jim Harris to write also a blogpost "Requirements, data quality and coffee"