Backlogs are a good idea. Backlogs ease the transition from the old “requirements up front” world to the new more dynamic agile world. Backlogs provide a compatibility layer for agile teams to interface to more traditional project management and governance. Backlogs even allow you to take a stab at done date!
Backlogs allow you to even out work between the quiet periods and the busy times. Backlogs give you a place to store good ideas which you can’t do just now. And because stakeholders can see their request is not forgotten they don’t need to shout for it today.
Yes backlogs are good. I’ve seen them work well myself and I’ve taught many teams to effectively use backlogs.
But – you knew there was a but coming didn’t you? – but…
Backlogs have problems, too many teams are labouring under the Tyranny of the Backlog, they have become backlog-slaves and practice something we might call BLDD – Back Log Driven Development.
(To be clear, when I say “backlog” I am primarily thinking of the product backlog – the long list of all the things the team (might) do in the future. This is different to the sprint backlog (iteration backlog). The sprint backlog is a shorter list of things the team aims to do this iteration. I am using Scrum terminology but the ideas are pretty much “generic agile” and I’m thinking more broadly than Scrum. Many implementations of Kanban feature a product backlog of sorts so while Kanban is less prone to these problem it is not immune.)
1) Lump of Work Fallacy
There is usually an assumption that the backlog represents all the work to be done – an impression reinforced by early implementations of Scrum. In the short term that leads to agile teams being seen as inflexible and prioritising process over need because new work is not allowed in.
In some cases teams even struggle to get started on work because a big-up-front requirements gathering and analysts activity is required to create a backlog. In the worst cases that work is even estimated and end-dates forecast before a line of code is cut or developers hired.
In the longer term it is simply unrealistic to assume the backlog is fixed. Even with more and better analysis it is impossible to foreseen future requests. The agile adage “it is in doing the work that we understand the work” cuts both ways: coders understand what they need to build and customers/stakeholders/analysts understand what they want.
Work will arrive after you begin, any system that does not incorporate that truth will fail one-way or another.
2) Bigger then you think
Not only does the backlog grow with completely new work the work in it changes – and grows. There are many reasons this happens: new opportunities appear, hidden ones become clear, requests require more work than expected and so on.
Humans are very bad at estimating, especially about the future, and, it turns out, they are also very bad at estimating time spent in the past. If you want accurate forecasts you need to invest in them, you need to make structural changes and you need to use statistics.
However, because of the lump of work fallacy and the belief that humans can make estimates, poor end-date projections get made and when they are missed (because they were wrong to start with) everyone gets upset.
3) Fallacy of Done
Backlogs come with burn-down charts and an assumption that there is an end; and that end is when everything is “done.” The team will be done when the backlog is empty. That assumption is baked into BLDD, traditional project management and even governance.
I have long argued that software is never done. I’ll accept that I might be wrong, but in the digital age, when business runs on digital technology (i.e. software) your products are only done when you business is done. The technology is the business, and the business is the technology. Stop the backlog growing, stop growing you technology and you kill the business.
4) Backlog Bottomless pit
Put all those reasons together and the backlog becomes a bottomless pit. In the early days of agile, when I managed teams myself, the backlog would often sit on my desk, written out on index cards and held together with rubber bands. I could get a sense of how big the backlog was my looking.
Today everyone uses electronic tracking systems. Not only do these allow an infinite number of items they rob us of perspective. To paraphrase Comrade Stalin: “2 outstanding backlog items is tragedy, 200 is a statistic.”
5) Backlogs obscure strategy & purpose
With so many backlog items it is easy to get lost – you can’t see the wood for the trees. Arguments over what will be done next start to resemble deciding who should get a lifeboat place on a sinking ship, add in the demands “when will you be done?” (plus explaining why the date has changed) and “the bigger picture” gets lost.
In Back Log Driven Development the sense of purpose and strategic goals is lost as teams struggle with the day-to-day demands of just doing stuff.
6) Powerless product owner (i.e. backlog administrators)
Tyranny of the backlog seems worst were product owners lack real authority and skills. They are little more than backlog administrators. They spend most of the week adding requests to the backlog, then passing a few chosen items to developers in planning meetings. A vicious circle develops, the product owner can’t win so people trust them less, their authority wanes, and the backlog spirals.
Few organisations give product owners the power needed to get a grip on this situation. Indeed, many product owners are plucked from the ranks for development or support and given a battlefield promotion to product owner but lack the skills required. (See The problem with Product Owners.)
For years I’ve been suggesting teams throw away the backlog – you will not forget the important things. But then how do you know what to do?
Take a step back, start with your purpose, your mission, the reason you team, your company, your organisation exists. What should you be doing? How can you fulfil that purpose and sever your customers?
This is where I see a role for OKRs and jobs to be done. Both these techniques – together, or separately – can be used as story generators. Every time you need to more work, more stories, you return to your OKRs and ask “what can we do now to move us towards our objective?”
When writing Succeeding with OKRs in Agile I became more and more convinced this is the path to take. Increasingly I sum this up as Purpose over Backlog.
Step 1: Clarify your purpose – what is your overarching reason for existing?
Step 2: Clarify how your existing strategy builds towards that purpose, and if you don’t have a strategy create one.
Repeat steps 1 & 2 annually.
Step 3: Think broadly, set your OKRs as a team so you build towards your purpose by following your strategy.
Step 4: Spend the next 12 weeks executing against those OKRs
Repeat steps 3 & 4 every 3 months.
Step 5: In each planning meeting take stock of what you have done and progress against OKRs
Step 6: Ask “what do we need to do next to move towards the OKRs?”
Repeat steps 5 and 6 every 2 weeks
And if you are Kanban’ing then keep steps 1, 2, 3 and 4, adjust 5 and 6 as appropriate.