Confidence in big data is highly variable. Some data sources have inherent uncertainty. So why shouldn’t you spend as much time as needed to make big data perfect? Time. You simply don’t have enough time to sort out every data irregularity, every ambiguity, every incomplete attribute. And for many big data use cases, you don’t need to. That’s why perfect is the enemy of good. In the era of big data, governance has evolved to first diagnose the usage, then prescribe the appropriate amount of governance. So the objective is not to make it perfect for every possible usage up front, it’s to make it good enough for the use case at hand.
Tony Baer of Ovum explains this in more detail in this blog post.
For more about confidence in big data, see my previous post, "Be Confident In Your Data."