Six Techniques to Split Large Stories
In part 1 of this series, we talked about the reasons to split large stories into smaller ones, even when it’s difficult to do so. You may be convinced that splitting up large stories is the right thing to do, but sometimes it seems impossible, no matter how hard you try.
In this blog post, we’ll look at six next-level techniques you can use to split up large stories. But first, let’s look at one approach that we would not recommend.
Not Recommended: “We’ll Do It As A Followup”
Often when working on a story you might finish the “main” functionality but having a number of “wrap-up” tasks remaining—for example, testing, accessibility, UI polish, or documentation. It can be tempting to move these tasks into one or more followup stories so that the main functionality can be reviewed and merged.
Be careful, though, as there are a lot of risks to deferring wrap-up tasks to future stories. If your tests really provide value in confirming the code works, then why merge the code before you have that information? If UI issues would prevent release, then merging the incomplete UI makes your application not releasable (unless that functionality is behind a feature flag). Also, even if you intend to work on these wrap-up tasks right away, priorities can always change. Keeping the code unmerged is an accountability to yourself that you have to finish those wrap-up tasks before moving on.
There’s also a deeper problem with deferring wrap-up tasks—doing so reinforces the mindset that a story is “done” when the main functionality is in. This mindset will tempt you to continue estimating and operating this way. To break this pattern, decide now that you won’t deliver this story until all the wrap-up tasks are done.
With that non-solution out of the way, let’s look at solutions that are recommended for splitting up stories.
1. Clarify Acceptance Criteria
When a story has vague acceptance criteria, it’s very likely the story will end up becoming large. Maybe there is a lot of functionality under the hood that is implied but not spelled out. In this case, every time you get something working you find something else that also needs to be built. Because you didn’t know about that work at the start to proactively split or sequence it, you may just continue working on it without thinking, or you may have coded yourself into a corner where it seems like you have to build it all at once. Avoid this trap by taking the time at the start to list out everything that’s necessary for a story.
2. Split Stories Proactively
If you approach a story by asking “do I think this is small enough or not,” you may be setting yourself up for failure. Humans as a whole are just bad at estimating software. Instead of relying on your questionable ability to estimate, try to split stories up as small as possible. Here’s the question Kent Beck asks in Extreme Programming Explained: “What’s the least you could do that is recognizably in the right direction?” If you find you’ve gone too far, it’s always easy to re-combine small stories into larger ones, but it’s hard to split large stories into smaller ones.
3. UI Component Library
For some stories, beyond the business functionality itself, a complex UI component also needs to be built. It can be a good idea to build and deliver this component in one story, then use it to implement the business functionality in a subsequent story.
One way to make it easier to deliver UI components as standalone stories is to set up an interface that allows you to see those components in isolation. This is sometimes referred to as a “living style guide,” and is a common practice when building a standalone UI component library. But this process can also add value if your components live directly in your application and if developing those components as standalone stories would help.
There are a few different ways to implement a living style guide. You could put it in an admin section of the app that isn’t accessible to end-users, or you could create a separate application or build a target to create the living style guide. In the frontend web world, Storybook is a cross-framework tool that can help.
4. Creative Ways to Set Up Data
Sometimes the need to set up data can make you feel like you need to build a lot at once. You have a user story to view some data, but your user account doesn’t have that data! So it seems like you need the functionality to create that data, even if the end-user doesn’t.
Before you decide to write code to set up data for yourself, there may be other ways to get that data set up:
- If there is another existing client application for your system that has adding functionality that you could use to create the data.
- Alternatively, you could ask a backend developer or database administrator to create records for your user account.
- Another option is to mock out the data by hard-coding it with a mocking tool like WireMock or by writing a lightweight web service. These options are more work to set up but can pay off if they’re needed for many stories.
5. Git Good
When you’ve put a lot of work into a branch and then discover that the story has become too big, what do you do? Leveling up your version control skills can help you reshuffle your work to support a smaller story. I’ll speak to
git here, but similar skills would apply to any version control system.
When you’re first building your story, create small, focused commits that only do one thing. Then, if you find you need to split out some of that work into a smaller story, you have a few options:
git commit --amendmakes changes to the last commit, allowing you to undo parts that are no longer needed.
git cherry-pickcopies an individual commit from one branch to another.
git reset [commit-hash]takes changes from a commit and puts them back into the working tree, to edit them before recommitting.
git rebase --interactiveis a powerful command that allows you to make changes to a series of commits, such as reordering, editing, or removing commits that are no longer needed.
To learn more about these commands, check out the “Rewriting History” chapter of Pro Git, available free online.
6. Feature Flags
Say you have a system in production that you’re making large additions or changes to. You don’t want to build the whole large feature at once, but you also don’t want to leave the system in a broken state by not shipping the full feature.
To accomplish this, you can use feature flags, which allow you to enable or disable features for specific builds of your app. This means that you can leave a feature disabled for production but enable it for developers and optionally QA engineers.
Feature flag setups can be very sophisticated, but if your immediate need is to split up a large story, I would recommend starting simple. Often an environment variable or build config variable is enough.
These six techniques can help you split a large story into smaller stories when there doesn’t seem to be a way to do so. They are useful in a variety of situations, but they won’t cover every possible project, team, and technology—you may need to come up with your own situation-specific techniques. Hopefully, these techniques serve as inspiration for the kind of creative thinking that can help you split your large stories so you can keep delivering value consistently.