This is because there's a special class of retrospective item that you can only reliably close by addressing it on a recurring basis.
New Year's Resolution? Lies, all lies.
For instance, suppose during your regular retrospective developers raise the issue that they don't know when to mark stories complete because they don't contain enough detail. Or, this problem could manifest itself as testers frequently rejecting vaguely written stories to the chagrin of developers.
In either case you could address the issue by promising to write stories with more detail or (better) to include acceptance criteria. Sounds great, the retrospecitve ends, everyone is happy since things will be much better going forward.
Only they won't. This solution is a mistake. Like most New Years resolutions, promises "to do better" are a lie. Promises like the above last for a sprint, maybe two. However, as time elapses and/or as stress increases, we forget our promises. We revert.
In the example above consider a customer verbally mentioning a new feature they want to you in a hallway. You quickly capture the need as you run off to another meeting with the intent to fill in the details later. An hour and four emergencies later you've already forgotten. The mistake wasn't breaking the promise, it was making it in the first place.
(Lists) to the Rescue
So how do we truly solve this issue? The answer is by adjusting our close-out process. Sprint close out is the perfect time to address issues like this because we can block out time to correct problems introduced during those hectic periods. Furthermore it's much easier to take a step back to focus on quality when you're not sprinting.
But if our processes are purely based on some static, published processes that isn't customized to our team and circumstances, or is based on the way we did things the last time, then we will lack the means to consistently and reliably adjust our processes over time. We need a different tool.
One such tool is a closeout checklist. The list may start as simple as listing the meetings from a traditional scrum process:
- Perform Sprint Review
- Perform Sprint Planning
- Perform Sprint Retrospective
But over time (more specifically after retrospectives) we can add items onto this list thus customizing it to the specific needs of our team.
In the case of the example above we can add an item prior to Sprint Planning such as "Clarify New Stories". We can keep the list simple or move it to a spreadsheet with columns like responsible team members (i.e. just the scrum master, develoeprs only, whole team), implementation details, due date, or date completed.
Sample Checklist Items
If this still sounds a little vague here are some sample retrospective items that I've gathered from my career and how one might adjust a closeout checklist to address the underlying issue:
|Post-production, stories are taking longer than expected||If they failed to take data migration into account, try adding an "identify migration issues" task to the checklist. Possibly limit it to more experienced developers to keep the meeting short.|
|The CI server takes too long to build||If there are lengthy UI automation or performance tests, consider moving them out of the continuous integration cycle into a nightly build. Then add a "check status of acceptance tests" task to the checklist. Don't close the sprint unless they are all passing.|
|Too many orphan branches in the source control server||Add a "delete orphan branches" task to the checklist, maybe assign it a specific person.|
|Lack of adherance to coding conventions||In Code Reviews: A Sane Approach I argue for a per-sprint code review. The checklist would be the perfect place to ensure this consistently happens.|
|Code Coverage is Low||Add a task to "Publish code coverage numbers" and include the delta from the previous sprint. Sometimes increasing visibility is enough to remedy issues like these.|
|Effort estimation takes too long||If eliminating it altogether isn't an option, try adding "auto-point defects to 1, no discussion allowed" in details of "sprint review"|
|Developer databases out of sync||You could solve the problem once, or better yet add "re-baseline databases with production" task to the checklist to ensure it doesn't happen again.|
So while a closeout checklist may sound pedantic at first it can be an excellent tool in your tool belt. If you give it a try, you may find that such a list can increase your maturity while becoming the linchpin that holds all of your processes together and ensuring that continous process improvement happens for your team today and far into the future.
So what checklist items would you add to the list above? Feel free to post in the comments or ping me on twitter.