Ed Pollack shares some great advice:
It is always an afterthought. New objects are created that start off small and current. New feature development takes over and the recently architected data structures become old news. Over time, data grows and suddenly a previously small table contains millions or billions of rows.
Is all that data necessary? How long should it be retained for? If there is no answer to this question, then the actuality may be “Forever”, or more honestly “No one knows for sure.”
Retention takes on many forms and this article dives into ways in which data can be managed over time to ensure that it is fast, accurate, and readily available.
We don’t tend to think about data retention in the development phase, but it’s an important consideration and thinking about it up-front might save you disk space headaches later.