What Do You Do When Work-in-Progress gets Cancelled?

Cancelled WIP

“Life is what happens to you while you're busy making other plans.”

(Lyric from “Beautiful Boy” by John Lennon)

Recently I was having a conversation with a client, and he raised a question that affects many teams - “What should we do when work-in-progress gets cancelled?” The underlying concern was - “How can we account for the time spent on these cancelled activities while still maintaining our predictability?”  It’s a really thoughtful question, so I wanted to capture some noodlings in this blog!

First - “Predictability” means something specific here

When we talk about predictability as a goal for a Kanban system, we're talking about process predictability - not task-level predictability in terms of hours, individuals, or estimates. We're interested in the predictability of the workflow, a.k.a. the delivery system.

System-level flow

We aim to understand flow at the system level, not at the level of the individual person working within the system. And we aim to understand the sources of system variability, knowing that there will always be some variability. The work of continuously improving system stability is really the work of understanding and then lessening that variation where possible - which is what leads to improved predictability over time.

Cancelled work is system variability

So where does cancelled work fit? It's noise in the system as the system chugs along, working to get items across the finish line. It's data about how the system behaves. And, not all cancelled work signals a problem! If work is cancelled because you've learned something about the customer needs or business opportunities which tells us we should no longer do this work - that's actually "winning." But if the cancelled work feels more like "churning" than "winning," a key question might be: how much cancelled work are you seeing?

Let’s say you start measuring cancelled work data and discover that something like 1%-2% of your WIP has been cancelled over the past 6 months, and that it happens at fairly consistent intervals. You might reasonably decide that this represents normal, acceptable noise in your system, and choose to focus your continuous improvement energy elsewhere.

 (Note that 1%-2% over 6 months is a made-up data point here for illustration purposes - your own thresholds will vary.)

On the other hand - if you start measuring and find that 30%+ of your WIP is being cancelled after work starts, that's a pretty major category of system variability!   Most importantly - that kind of insight would be a data-driven invitation to dig deeper. 

If you dig deeper and determine that your cancelled work situation is recurrent and significant, perhaps you decide to shape up a workflow improvement experiment targeted at reducing the likelihood of cancelled WIP.  Why is work getting cancelled? Are there any themes? If there are themes - which interventions or changes might help?

For example, teams I've worked with in the past have made meaningful improvements in situations like this by introducing new workflow policies aligned with recurring themes, such as:

• Work should be confirmed as a priority before starting

• Acceptance criteria/definition of done for this work item should be approved by key stakeholders before starting

• If this work requires collaboration with others outside the team, those people have confirmed they're available and ready 

• The work item is right-sized to flow through our system quickly - it is small enough that it is unlikely to be overtaken by changing priorities before completion

So.. should you account for cancelled work? Yes - but for the right reasons

The goal of tracking cancelled work isn't to account for where people's time went. It's to understand a category of system variability so that you can inspect it, learn from it, and potentially improve it.

This is where the four foundational flow metrics - WIP, throughput, cycle time, and work item age - can act as a North Star. They help keep the focus on flow at the system level, even amid all of the glorious and unique types of variability you'll see across different teams and workflows! Cancelled work shows up in that picture: it ages, it consumes WIP capacity, and it shapes your throughput and cycle time… even when that work never crosses the finish line.

Cancelled work: Track it. Look at it regularly. Let the data tell you whether it's a signal worth acting on or acceptable noise to live with for now.

This is the beauty of a data-informed, continuously improving system. You don't have to have all of the answers up front. You just have to be willing to keep looking.

Keep looking and continuously improve