Originally Scrum was very much about “You tell us what needs building. We’ll decide how we build it and how soon we’ll deliver.”1 I’ve never seen that version of Scrum. The version I have seen, has a product manager try to get as many features into a sprint as reasonably possible - for varying degrees of reasonable. This comes at the expense of maintenance work, such as keeping libraries up-to-date or removing technical debt. And it incentivizes the team to cut corners on features, to not leave code in a better state than they found it, to not fix smaller bugs and instead log them somewhere for later.
One solution I see to this problem, is to put an engineering manager fully in charge of the team.2 The product manager prioritizes the features. The engineering manager prioritizes the full scope of work for the team. That’s not a simple change to pull off, however.
Another solution might be to change the way we use our backlogs. If a product manager gets to prioritize all the work, and the tool they use is a backlog, then we should make sure that all the work3 is in the backlog: features, bugs, and technical debt. Let’s take a look at each of these three categories of work.
If you have a backlog, it already contains features. So no need to change anything here. You could argue - as I did in “Our work management tools are limiting our imagination” - that there might be better ways to manage work than having features in a backlog. That’s not what this post is about. The assumption is that you have a backlog, owned by a product manager or product owner, containing the upcoming work for the team.
Putting bugs in your backlog is a tricky proposition, because it becomes the default action for bugs that are not obviously urgent. So before you know it, you have a bunch of bugs in your backlog and all of them feel kind of important but also not that important. The number of bugs grows, some of them become outdated, it’s too much work to re-triage them all again, and now what?
A solution to this problem is to first ask: Should we fix this immediately? With immediately meaning either that someone drops what they’re doing and starts fixing it right away, or that someone wraps up what they’re doing and then picks up the fix. For a bug in a story that’s in development, in most cases the answer to “Should we fix this right now?” should be “Yes.” Fix it now and you’ll never have to think about it again. For other bugs, e.g. based on a support ticket, the answer will be “yes” less often, because they’re not related to the work the team is doing at that moment.
If you decide the answer to the question of fixing the issue immediately is “no”, it’s important to be very clear what that means. Often that “no” seems to be taken as “Good thing we logged it, so at some point in the future we can get back to it and fix it.” While that comes from a place of good intentions, I think there’s a better way to look at that “no”. By not fixing it right away, you accept that the issue is now part of how the product works, perhaps indefinitely. And doing the fix has become one of the many items in the prioritization cycle. And there’s no way of predicting with absolute certainty when it will get enough priority to get picked up. So it might never get fixed at all.
That leads us to the third option of dealing with issues, simply forgetting about them. You’re not fixing the issue immediately. You’re not fixing it soon, so it shouldn’t get a spot in the backlog. Don’t weigh yourself down by keeping track of them, let it go.4 And for those contexts in which there is value in logging all bugs, whether you will fix them or not, do log them, but not in your backlog. Record them somewhere else.
Technical debt and maintenance
“Technical debt” has many different definitions, with plenty of disagreement in which is the correct one. In this post I’m using the term in its more general sense: “tech stuff that’s making us slower than we could be”. A related type of work is maintenance work, such as keeping dependencies up-to-date. For the purpose of this post, they both fit this third category of backlog items, all the work of which your engineers are the main users and stakeholders.5
While features and bugs tend to get added to the backlog, technical debt and maintenance usually don’t. Or at least not in the same way, the bar seems to be significantly higher. There’s a reluctance.6 They only get added to the backlog when the work really needs to get done soon. And something I have yet to see, is a team consciously incurring some technical debt and immediately adding an item to the backlog to pay off that debt.
So what you tend to end up with is a product manager or product owner with an incomplete view of technical debt and maintenance work, because:
- They never interact with the codebase that lets them experience how tech debt slows down the work. It’s an annoying fact-of-life in software development that explains why things take longer than anticipated. It’s not something you can actively manage.
- The tool they use to manage the priorities, the backlog, is ‘complete’ when it comes to features and bugs, but not when it comes to technical debt and maintenance.
The first of these is definitely worth doing something about, e.g. through ensemble programming (also known as mob programming). A way to address the second one is to add technical debt and maintenance to the backlog in exactly the same way as features and bugs. Scrolling through the backlog should give an accurate picture of the actual ratios of the three types of work. Make all the work visible.
All backlog items are equal
Features, bugs, and maintenance all are things we could do to increase the value of the product. (Or at least, we believe they would increase it.) In that sense, from a work management perspective they are the same kind of thing. They are all backlog items. Work you intend to pick up soon.
If you want to, you can give them different labels, but I think there’s a misleading aspect to doing so. At what point does a bug become a feature? A wrong calculation is clearly a bug. But what about supporting part of a process better, making our users more effective? And why do we treat the people working on the system, as second-class users? Why are the features and bugs they care about, treated as lesser than those of other users?
Hence my proposal for a simple experiment. If you have a product manager who gets to make all the priority decisions about the work, then put all the work in the backlog: features, bugs, and maintenance. Make it visible. And if someone complains this makes the backlog too large, then have a conversation about what can be thrown out. But at least you’ll start that conversation with a more complete picture of the work that could be done to increase the value of the product.
Thoughts from other people
This blog post started as a question on Mastodon:
“What if you started to log technical debt as bugs in Jira?”
In response, Krys said:
“I fear relabeling would only reinforce everyone’s troublesome view of the PO being the teams boss and having authority/obligation to micromanage their workday instead of demanding technical excellency from the team while trusting them to know best how to get that”
Markus Tacker said:
“Nothing. Technical debt is valid implementation. As long as developers do not get permission (or sense of urgency) to change it, it will get ignored. Work on providing the resources needed to address it instead.”
“Tech debt impacts the ability to deliver new features. And that’s why it does not belong in the backlog. Only features do. If feature backlog is empty, no need to address tech debt. If feature backlog keeps growing: fix tech debt.”
(edit 7 April 2023) After publishing this post, I learned that about a month earlier Dan North had published “But what about the BAU work?” and that a day after my post Martin Fowler published “Slack”. Both articles highly recommended!
A solution I don’t believe in is that engineers should learn to have better conversations with their product manager or product owner. “Having better conversations” tends to mean that engineers needs to do a better job explaining the business value of the technical work they want to do. This keeps product management in a position of power-over instead of power-with engineering, where any conversation about what to work on, needs to happen on product management’s terms. ↩
Sometimes a cheap way to log a bug is adding an automated test that captures the bug’s behavior. Do make sure it’s not just as easy to also fix the bug, though. ↩
I suppose that makes compliance-related work either features or bugs? ↩
There’s also a reluctance in getting this work prioritized. For the non-urgent work, it often ends up as filler, when the product manager can’t find anything else for the team to do. ↩