Prove
It is essential to confirm the new system is delivering the expected results. This may be assessed by manual means like simply entering test data and checking the results. It may be that a series of tests are completed regularly by automated means such as sophisticated unit testing.
Regardless of the means, be it manual or automated or a combination of both, you need
Thorough Testing
In general there is no substitute for thorough testing in a controlled environment before rolling a change out to the entire business.
It may be appropriate to create a duplicate version of your live system and conduct all testing off-line. For smaller changes it may be practical to simply have your Project Champion alone work from the new version and then isolate and review their completed transactions.
Do we rely heavily on appropriate testing? Yes. We believe that an hour of testing regularly saves us many hours of unravelling and backtracking! The bottom line is it is commercially unwise not to have a level of testing and confirmation in place that is commensurate with the risks and costs of failure.
Fallback Planning
Thorough testing usually ensures all the obvious issues are addressed before the system goes live. But no testing is 100% perfect. It is also prudent to determine a suitable fall-back strategy in case a problem is encountered. This may be as simple as a quick roll-back to a previous stable version of the application if a bug is identified.
Do we find that we rely heavily on our fallback measures. No. They often go unused. And they can seem a burden, but only until the first time you need to use one! So again, it is commercially unwise not to have a level of fallback planning and disaster recovery in place that is commensurate with the cost of the potential disruption to operations.
Birthday presents are fun but surprises in software are generally undesirable. It is important to have considered the risks and formulated an appropriate testing and response plan.
