Saving history is crucial
A data platform that does not store its input is a very fragile platform, as it needs all the input systems to be available at the same time of each run to be able to produce its results.
An even bigger limitation is that a platform without history cannot fulfill many of the requirements that are otherwise possible and expected today by a modern data platform, such as auditing, time travel, bi-temporality, and supporting the analysis and improvement of operational systems and practices.
To us, anyway, the core reason why you should always save the history of your entities is to build a simpler, more resilient data platform. When you split your platform into one part that just adapts and saves the history (the storage layer) and another that uses the saved data to apply the desired business rules (the refined layer), you apply the principles that we have discussed and achieve a great simplification of your data platform.
Figure...
