A couple of years ago, the team I was in started to use a concept we called gradual refactoring, whereby we would allocate a fixed percentage of our capacity to doing refactoring in order to keep technical debt at bay. That worked pretty well, and has been refined over the years into something we now tend to call the Productivity Backlog, in contrast to the normal Product Backlog. The way we use this idea boils down to reserving a more-or-less fixed percentage of overall capacity to improving that capacity. This is typically refactoring or cleaning up old, hard-to-change code or building or improving tools to help with something that slows us down. Normally those tools help us automate some process, making it less error-prone and less labour-intensive.
This works really well. Initially, I wanted to keep productivity efforts as part of the normal backlog, because I wanted a global prioritisation order. “All we need is to get really good at weighing medium to long-term benefits of productivity improvements vs the shorter term benefits from the product improvements”. That has turned out to be even harder in practice than I had thought. It’s just easier to keep things separate, for the following reasons:
- It’s an easier sell to the non-technical parts of/people in the organisation. “Continuous improvement” is something almost all organisations would like to do, and keeping a Productivity Backlog is an easy way to formalise that at a team level.
- It means that, as a technical person, you don’t have to try to sell the benefits of intangible things like “refactoring” or “eliminating technical debt” each time you need to untangle something particularly troublesome in your code base.
- It means that, as a business person, you don’t need to try to understand the reasons why some code is poorly suitable to some task or other, freeing you up to focus more clearly on figuring out what features and changes will make the product more successful in the future.
- It simplifies prioritisation both on the product and the productivity side, because a) you are comparing apples with apples (or at least fruits with fruits, if you know what I mean) on both backlogs and b) you can have different owners who can structure and order their own backlog without a need to coordinate needs and requirements from the other one.
- It provides an easy knob to turn as needs change – if you urgently need to get some product feature out, you can easily turn the productivity backlog down to 10% or even 5%. And if you’re suffering acutely from wasted time due to inefficient deployment procedures, you can turn it up to 20-25% until you’ve solved that crisis. If you do things on a case-by-case basis, you have to weigh the benefits of saving 3 minutes each time somebody needs to deploy a particular build of something somewhere against trying out a new front page to improve conversion, something that is hard and requires people with different types of expertise to understand each other’s problems and agree to a compromise.
Some of the issues we’ve had in using this concept have been around figuring out what types of changes should go on it. Initially, we called it the ‘tech backlog’, because it was owned by ‘tech’ and dealt very much with issues in implementation. That led to attempts at adding things like performance problems/measurement, or bug fixes driven more by technology than by the business side of things (like things getting logged that made real problems hard to find even though there was no obvious end-user impact). But those things are product features – quality and performance are clearly features in the product. Productivity is about the team’s ability to deliver more product features over time. That means it can be about fixing bugs, if those bugs, for instance, mean that deployment occasionally fail. But most of the time, it’s about removing obstacles and frustrations that keep us from delivering the product improvements that are what we get our job satisfaction from.
Productivity backlogs, compared to product backlogs, are reactive more than proactive. You feel the pain of some faulty process or code, and when it grows unbearable, you add fixing it near the top of your productivity backlog. There’s very little point in having long term goals for a productivity backlog.
A separate productivity backlog makes life simpler for both the business and the delivery side of things, is a great way to formalise continuous improvement and helps improve job satisfaction in delivery teams partly by removing obstacles to productivity and partly by providing knowledge (or hope at least) that the muck you’re having to wade through at the moment will at some point in the future be cleared out. Everybody should have one!