Anyone that’s worked in a genuine continuous delivery environment will find it very frustrating when features are completed but not released to production. Abuse of feature toggles can make this worse – blurring the line between what a team considers complete and what has actually been released to a real user. We know that when you actually release features with a big delta to the wild you will likely be faced with various bugs and user issues which will need to be fixed and polished to make it really complete – and the bigger the delta the worse they will likely be… In the worst case you may be building something that is completely irrelevant to your users (which could have been discovered many sprints previously).

Anyway, regardless of all that these thoughts remind me of my time many years ago working at Westfield Digital when we would do a 2 week sprint, followed by a long cycle of test and regression before the features actually made it into production. The focus was always based on how many points were delivered for ‘completed work’ rather than ‘released work’ and I always felt this had an effect of the quality of the work completed each sprint.

The following cartoon is a shameless tribute/rip-off of XKCD that I did ages ago thinking on this.