The Problem: Confusion, Uncertainty, and Delays
We find this a common thread amongst our customer teams, which can be attributed to a lack of backlog refinement.
“What do you mean we are changing the way we list items? This is going to be messy – does finance know? Well fine, we can do it, but it’s not on me when we have to roll back.”
Sound familiar? When the detail of what needs to be built gets decided too close to the development phase, the integrity of your flow of requirements from idea to reality becomes sensitive to small events. The above statement is paraphrased from a real developer on a team building a financial research product. Someone was out of the office for one day, and a decision was made in their absence because “we needed to get on with it”.
“It's the end of the Sprint but these features are not ‘done done.’ I guess we’ll have to carry them over (again).”
It is perfectly normal for a team to miss their forecast now and then. If it happens most times, there is a very good chance that the following pattern has been established:
- Sizing is done during Sprint planning.
- There is no time for the detailed task breakdown or solution design that can help teams forecast effectively.
- The team may not feel accountable for improving their forecasting and refinement practices.
- Planning is painful and cyclic because no one understands the stories.
“We forgot to submit this feature for SecArch review, so we’ll have to postpone it for a couple of Sprints.”
External dependencies increase complexity and ideally development teams should have the skills and mandate to handle architecture, design, and security reviews themselves. If you are not there yet, a mitigation technique is getting clarity over what is coming and what needs to have been done by what stage.
The Solution: Backlog Refinement
In other words, detailed business analysis, documented well, organized effectively and executed at the right time. Backlog refinement is a general term encompassing all of the analysis that takes place before the work is started. This might include:
- A regular meeting with the whole team to estimate features and agree on acceptance criteria.
- Smaller groups working on test plans, architecture, and design.
- Individual specialists preparing information to be discussed with the team.
- Anything that helps the team plan and forecast effectively before they start the work.
This is bad news if you interpreted the line in the Agile Manifesto “individuals and interactions over comprehensive documentation” to mean “just make it up on the fly without having to come to any boring meetings.” The level of detail required to bring a successful product to market is broadly the same whether you employ traditional or Agile methods; it’s just a question of timing. In Waterfall, all of the analysis is done up-front, which takes time and makes responding to change difficult.
Conversely, taking an Agile approach means the detailed analysis is simply deferred until there is a commitment to build each feature. This takes inspiration from the dramatic productivity gains achieved in lean manufacturing piloted by Toyota in the 1950s. Here, “just in time” delivery of components reduced warehouse costs and prevented waste, as inventory become obsolete. The trick is to tune your system to establish a buffer so you don’t run out of work and set that buffer at the lowest safe level.
Set Your Product Backlog Refinement Buffers
Before work begins on each feature, it’s likely that the following has happened:
- A high-level description (such as a user story) has been written.
- The team has sized/estimated the effort required to build it.
- Acceptance criteria have been written.
- Supporting documentation has been created (wireframes, architecture diagrams, specifications, etc.).
Make your own list of these types of analysis, and note down when you currently do the work. For many of the teams we have been invited to help, it looks something like this:
Review of Timings
We find that considerable improvements may be made by simply reviewing these timings based on the lean principles that Toyota proved so effective. In lean, this became known as deciding as late as possible and was later codified to be a fundamental lean principle. The last part of that phrase is crucial: “as possible” – or to expand this a little, “Decide as late as possible without compromising your ability to do the job effectively.” The advantage of deciding later is you are not constrained by your choice. For example, if you have 6 months worth of wireframes, then changing the plan would involve throwing away a considerable amount of work. To find when “as late as possible” might be, ask yourself when the last responsible moment might be. Here is an example of how this worked out at one organization:
This table worked for one organization, but it will not work for you.
An Agile organization thrives on inspection and adaptation of the work being done and, crucially, the process of doing it. If you want to reduce confusion, uncertainty, and delays in your team, then a good start would be to theme a retrospective around backlog refinement with an agenda as follows:
- Make your own list of requirements analysis and add in anything you think might be missing.
- Get together as a team and try to guess what the last responsible moment might be for each form of analysis.
- Implement your schedule by regularly holding refinement sessions to get each backlog item up to the level of detail required.
- Assess the results regularly during retrospectives. If you think you are analyzing something too early or too late, change it!