Making System Interventions with Kanban Thinking

This post pulls together a number of ideas on interventions into a single place, and will become the content for a page on Interventions on the Kanban Thinking site.

What is an intervention?

An intervention is an act of becoming involved in something in order to have an influence on what happens. It is an active and participatory action, as opposed to a passive and observational instruction. Thus in Kanban Thinking, interventions are ways of interacting with a kanban system with the intent of improving it and having a positive impact.

There are 5 types of Kanban Thinking intervention, which can be considered as heuristics to help the discovery of appropriate practices and techniques for a system’s context. These heuristics may point to the purpose behind known and proven practices, or they may lead to the identification of new and alternative practices. They are:

• Study the Context
• Share the Understanding
• Stabilise the System
• Sense the Capability
• Search the Landscape

Study the Context

What could be done to learn more about customer and stakeholder needs, the resultant demand and how that demand is processed?

Kanban Thinking is based on a philosophy of “start where you are now” and the foundation for this evolutionary approach is Studying the Context. In his book “The New Economics”, W. Edwards Deming famously said, “there is no substitute for knowledge”, and it is the acquisition of this knowledge about the current situation that leads to an appreciation of how to improve it.

Various practices are widely used to study different aspects of the system. Empathy Mapping is one way of learning more about customer and stakeholder needs. The work that they request as a result of those needs can be studied with demand analysis and profiling to determine classes of service and response. The way that work is then progressed, from concept to consumption, can be examined using forms of value stream mapping.

Share the Understanding

What information is important to share, and how can tokens, the inscriptions on them, and their placements, provide a single visual model?

Sharing the Understanding of the current situation, typically in the form of a kanban board, creates an environment where people are more intrinsically motivated. The ability to see what needs work provides autonomy, the ability to see where improvements can be made provides mastery, and the ability to see where to focus provides purpose.

Potentially, lots will have been learned from studying the context, so it is important to decide what is the most relevant information to share to avoid being overloaded with too much noise. That information can then be visualised using various patterns based on the acronym TIP; Token, Inscription, Placement. Tokens are the cards, stickies or other tangible representations of the work, where aspects such as shape, size, colour or material can represent different information. Inscriptions are items of information added onto the tokens, where words, dates, pictures or symbols can all be used with suitable formatting. Placements are the Tokens are placed on the board to convey information, where horizontal, vertical and relative positioning can all be meaningful.

Stabilise the Work

What policies could help limit work in process, and remove unnecessary or unexpected delays or rework?

A kanban system which is stable is more likely to be able to evolve and have resilience in the face of change. Systems can be considered to have boundaries or constraints, and those with too loose constraints will devolve into chaos, whereas those with too tight constraints will become brittle with bureaucracy. Stabilising the Work is achieved with enabling constraints, open enough to avoid restriction and prescription, yet limiting enough to stimulate expansion and new possibilities.

Defining work in process limits is a primary means of stabilising the work. One approach is to start with the current work in process (WIP), and look to gradually reduce it. Alternatively, the WIP could be drastically reduced to stabilise quickly, and then gradually increased to a more optimal level. A middle ground could be to base an initial WIP in the size of the team, such as one work item per person. How WIP is spread across the system is also a factor, from a single limit covering everything (CONWIP) to a different limit per stage or column.

WIP is a form of explicit policy, and other common forms are Definitions of Ready and Done and similar, simple quality criteria. Additionally, Test Driven Development can be considered an explicit policy on how software is written to create a stable code-base.

Sense the Capability

What measures and meetings might create insights and guide decisions on potential interventions?

As well as acquiring knowledge on what the work is, and how it gets done, it is also important to know how well it gets done. Sensing the Capability of a kanban system involves getting a feel for its performance by both measuring and interacting with it. Metrics provide an objective, quantative sense of capability. Meetings, and other feedback cadences, provide a subjective and qualitative sense of capability.

Measures should be derived from the desired impact, and the outcomes which would make the impact. If Flow is the desired impact, being more responsive might be a good outcome, and thus Lead Time might be a good measure. Additionally, the subsequent behaviours, both good and bad, which a measure might drive should be considered, and trade-offs can be made by having a set of balancing metrics. A focus on Lead Time might result in a reduction in Quality by cutting corners, so measuring Released Defects could help balance that trade-off.

Meetings provide a cadence with regular interactions can generate feedback. A simple metronomic cadence can tie various events such as planning, reviewing and retrospection to a single rhythm, or a polyrhythmic cadence can decouple these events into multiple rhythms. Another option is for some events to be asynchronous and triggered by the work rather than time-driven.

Search the Landscape

What small experiments could be run to safely learn the impact of different interventions?

Searching the Landscape of a kanban system involves looking at a range of possible interventions and experimenting to discover what impact they have. Those that have a positive impact should be pursued and amplified. Those that turn out to have a negative impact should be reversed and dampened. This searching can be thought of as exploring the evolutionary potential of the present state, as opposed to defining a future state and trying to move towards it.

Defining an experiment involves proposing a hypothesis that has a rationale behind it, can be measured in order to validate or falsify it, and can be easily recovered from in case of unanticipated results. The intentional act of documenting the experiment, using formats such as A3s, encourages disciplined thinking and avoids falling into the trap of retrospective coherence to explain results.

Searching the Landscape is an exercise in continual curiosity about how to evolve and improve the kanban system, increasing impact through further insights, interactions and interventions.