Unit testing and integration testing have become part and parcel of development processes the world over. So why don't we apply the same rigour to the data persistence layer?

We trust that our unit tests will catch bad data being sent to the data store, but what if they let us down? A great (or terrible, depending on your view) example of what can happen is the failed Horizon project at the UK Post Office.

I'll propose why we should, and how we might go about adding a robust testing mechanism for the data store which can help spot issues earlier in the SDLC before end users are involved and start reporting issues.