One of the most popular phrases in the Data Profession is "a single version of the truth" and it is something that, over the years, has begun to grate on me and many others. Aside from the fact that experience tells us that perspective and objectives play a massive role in what people consider to be 'the truth', the pursuit of a single version often leads to a expensive centralisation of data to achieve a goal that seldom yields the outcomes we seek. To use Nick Vivion’s article, by proclaiming that we can have a ‘Single Version of the Truth’, we essentially confirm that there are multiple versions to begin with, and as such this cannot be truth but opinion. As technology and its capabilities have rapidly evolved, making the accessing and using data easier and simpler, the need for the mythical 'single version' becomes less important. Instead, the significance of well defined, modelled and governed sources where key attributes are defined, licensed, managed and owned, is far more important to the consistency of data products and the insights they produce. It may be time to change the message? #DataFabric #RedefineSource #Data https://lnkd.in/etaaf7vP
100% (ie true) . The “single source of truth” has some surface appeal, hence much selling of that ideal, but technical solutions aiming to achieve this are usually just aggregors or copies. This is a lossy or costly process. The aim should be the opposite- to get a better understanding from our ability to access more and more sources, in a quality way.
Interesting Matt Clark but what I think is most interesting is the interpretation of single version of the truth. I've never believed it means all the data in one place. To me it has always been about locating the true root source of the data & ensuring appropriate processes are in place to maintain that source as a single version of the truth. Just thinking tills & finance systems - in some companies tills are king for HMRC purposes. In others they're not because of the ways inter brand discounts are dealt with. I think a 'single version of the truth' is simply a way to get people behind understanding the importance of clean & accurate data. But then that's my interpretation of the truth!! I'm currently going through a process of trying to get 'one version of the truth' - but that's only for the people, asset & location data. I wouldn't dream of having that in with everything else. I think that's where you clever data architect people come into play. I'll tell you what I want, why / what I'm using it for - you work out what the root of that data is & get it to me in the most efficient usable way (hopefully highlighting any flaws in the business process) Simples!
Interesting perspective! Focusing on well-defined and governed data sources does seem more practical than striving for a single version of the truth.
Agreed and I will stop saying it, and update my terminology
Thank you Matt Clark raising this key and frequently misunderstood concept. Truth or veracity has context. There is no 1 single context. There’s also the element of time. Context can change with time as other factors as well. So expecting a single version without taking into consideration context and time is inherently flawed.
Single version of the truth works for top level corporate kpis but little else. Particularly at an operational level there can be many equally valid versions of the truth depending on the perspective of the person using the data
Good article and informative. As I always guide my clients, “Trusted source” is more relevant than “single source” !
Love this - well put Matt Clark
Very informative. Thanks
Head of Solutions Architect at Sky
5moMaybe it’s the same thing but a different way of saying it!