Data scientists seem to get lost in the idea of data normalization when the real issue is standardization. For example, if you want to build a successful business and then take that success to a second business then you probably need your tools to come along. But if the systems that feed those tools are faulty or represent something different then you've failed.
One of my clients has revamped their systems multiple times over the years and the problem is that there are no measurements to see if the changes had their desired effect. And that measurement is impossible now; while we are dealing with apples and oranges there is no way to make juice; which could be compared.And the real challenge is that at scale a single record might be as small as a grain of sand on a very large beach and yet doing something with that data and giving it the attention that it needs is complicated at scale.
It's that reason why I do not want to be a google size company.