You Guys, Data Is Really Hard

I've been part of multiple S/4 transformations. You've got your typical big project phases: blueprinting, analysis, build, and so on. Some of the toughest work of the project happens in the integration testing cycle(s). Scripts have to be written, authorizations handed out to the right testers, code and config transports properly moved, and…just so much crap to do.

When you execute the test phase, you'll accumulate a heap of defects. If your experience is like mine, even a cursory glance at the pile should make you shout "omg our data is F--KED". There is so much time wasted moving missed or mangled data, which brings important testing to an agonizing crawl.

To the best of your ability, determine what the golden truth of your data is and RUTHLESSLY compare the systems in your environment to it. Figure out what is missing from where it should be, what is present where it shouldn't be, and the minutiae of field-level comparisons. You might even have to write special data validation programs (I've been there). The first few times you do this, you will probably want to scream. This is understandable - it's painful. But if you push through, tune your processes, and painstakingly maintain your migration scripts, the sun will begin to peek through the clouds.

A lot of data stories are about feeding your enterprise AI beast, or your analytics strategy, or whatever fancy data-buzzword-of-the-week. But the only way to get to Data Buzzword City is by going through Boring Data Operationsville. And if you don't maintain the roads in Boring Data Operationsville, you'll never make it. PM