Kanban talks about limiting work in progress (WIP) as a way to manage workflow. “Limiting WIP is hard enough, but selling it can be nearly impossible” said Jim Benson. At the Lean Kanban Central Europe 2014 Conference he talked about how to convince others to limit WIP.
Knowledge work is invisible and adding new work has no apparent costs. We want to limit WIP to focus and prevent being flooded with work. Healthy work needs healthy constraints so that we can do the right and avoid costly coping behavior said Jim.
Although limiting WIP is obvious, people are not sufficiently aware that they need to do. Thus they are not looking for ways on how to do it. According to Jim they need an epiphany but they don’t want an epiphany, so you can’t give them one. What you can do to break the status quo is to show to people how it can be done.
Jim presented the five steps to an epiphany for limiting WIP:
- Acknowledge current state - define status quo
- Envision future state -define potential future
- Hypothesize active system elements - what fuels your disruptive systems
- Show don’t tell - put elements in action with visual systems
- Sell, baby, sell - find unobtrusively ways of highlighting positive messaging
InfoQ interviewed Jim about the importance of limiting WIP and what makes it so difficult, convincing people to limit WIP, and using measurements to manage workflow.
InfoQ: What makes it important to limit work in progress?
Jim: Limiting WIP without understanding the work itself means you might work faster, but not actually understand the work your doing. Therefore, it's important to see your work and control your WIP.
The Need to Visualize Work - We can better manage what we can see. Whether it is for oversight, team planning, or completing our own work, keeping the work visual means we can examine options, prioritize, see where bottlenecks or work starvation is occurring, and understand how work is actually completed. Visual information compels people to act. Problems are solved more quickly.
The Need to Control Work in Process (WIP) - When we overload any system, quality and production suffer. When we overload people with work, completion slows and defects rise. Rework increases, which increases load on the system. Most knowledge work projects suffer from overload and its negative impacts. Controlling WIP is a lever we can adjust to balance workflow with quality.
Understanding Planned and Unplanned Work - Visual systems with controls engender healthy production units. Previously invisible work becomes observable. When we can observe what is happening in our own work, or that of our teams or company, we can see where unplanned work is introduced into the system and its impacts. Note that all rework is unplanned. The class will show how to measure unplanned and planned work and make smart judgments about how to balance the two.
Using Personal Kanban Techniques to Understand and Control Work - Using a Personal Kanban board, an individual or a team can track work in all phases (not yet done, the various in-process phases, and complete) ... seeing work through its lifecycle gives us a massive amount of internal business intelligence: how we learn, how we prioritize, what impediments we find, where we can improve, when we can effectively collaborate, and more. In essence, we currently have no way to see what we are doing. We currently manage our projects as if we were driving our cars with our eyes closed. We want to use the Personal Kanban to open our eyes.
InfoQ: Limiting the work in progress sounds simple, yet many struggle in doing it. Why?
Jim: There are a couple of reasons what makes limiting WIP more difficult than people initially expect:
Options and Variation - Limiting WIP is difficult in a high variation environment. When conditions change, the next immediate step can be difficult to identify. We have many options. Selection is difficult, therefore we end up doing several things at one.
We Are Not Alone - Our coworkers are part of our team. They have needs. We have information. We end up being attached to work initially defined for other people. Since we usually don't understand our work, we also don't understand collaboration. That means that rather than having a system for coherent collaboration - almost all collaboration seems like or even is an interruption. Interruptions are unplanned and often untracked WIP.
We Limit WIP to 100% Capacity - People figure "my capacity is 3" and limit their WIP to that. Then it is frequently blown because we have no slack for unplanned work.
External Actors Seem Unjust - People outside our convenient planning bubble have needs. Again, we have planned our time to allow no slack and we are surprised when we are pulled in by clients or other forces of nature. This can include your boss.
InfoQ: Can you give some examples of how you convinced people to limit WIP?
Jim: I usually convince people by showing in stead of telling. You need to show people how and even why to control their WIP. My partner Tonianne and I will tell people to limit WIP, but by and large we need to build a visual system to show them the impacts that overwork has on individuals and teams. We can do this by simply showing the avalanche of work, the cost of delay, the level of existing rework, the amount of defective product being released, or even the morale of the team. We try to find external examples in their own organizations to make our point for us.
InfoQ: Which benefits did they get from it?
Jim: Shorter cycle and lead times (work done faster), lower escaped defects (work done better), better reactions to changes in context (work done smarter), and happier team members (work done healthier).
InfoQ: Measurements can be useful to manage ongoing work. Do you have some tips how to use them effectively?
Jim: We tend to rely heavily on numbers when we talk about measurement. This has led to management-by-objective, balanced scorecards, or other dehumanizing techniques that use proxy metrics rather than first hand observation.
If I were to ask you, "When you eat dinner, what forms of measurement are you using to know if the food tastes good or that the fork makes it to your mouth?", you'd think I was crazy.
There are extremely useful real-time data that we gather as we undertake a task. The most effective measurements we can use to manage on-going work are simply to observe the work as it is occurring, to see the flow of work, understand what decisions are being made, and to react as quickly as possible to changes or issues. We want to use real-time or leading indicators to spot future defects and correct before they happen.
Some usually overlooked generators of measurements are:
Cognitive Bias - Because knowledge work largely happens in the minds of individuals and is reliant on constant group or individual decision-making, we fall prey to cognitive biases and other human foibles. Decision-making short cuts that are natural for the brain but detrimental for business are common. The Planning Fallacy, Experimenters Bias, Fundamental Attribution Error and many others often result in poor planning, faulty assumptions, and costly unfounded expectations. Visualization of work and control of WIP create systems that can highlight these issues and allow for proactive correction.
Subjective Well Being - SWB is the monitoring of team or individual worker health simply by tracking mood. Difficult to do without a visual system, this is trivial with a Personal Kanban. Companies like JP Morgan and Cisco use their Personal Kanbans to track how staff feels after every task. If morale seems low or in danger, managers immediately seek to find out why. Low morale is a leading indicator of quality issues. Simply put, happy workers do better quality work. We want to track that and respond when necessary. Starbucks would never tolerate an angry barista because their mood directly relates to customer satisfaction. It is no different for staff in their headquarters.
Having said that, metrics for each individual project are very important. Choose metrics that give you only information you need to know. My short list are cycle time, subjective well being, and throughput. With these numbers, we can gather enough project-health information to know how long customers should wait for product, how much we are completing at the moment, and how our team feels about the project and their work.
With small batches, we should raise the number of releases so that we get client feedback as soon as possible.