Kanban Flow Physics

Our goal is to deliver maximum value at minimum cost. We seek to achieve this via a relentless focus on Cycle Time reduction – Cycle Time is a proxy for cost. The cost to produce an item is directly proportional to the amount of time it spends in the system. Thus, reducing Cycle Time reduces operating costs.   The organizing framework we use to achieve this is the Kanban Method. The Kanban Method is a set of principles, methods and tools that enables an organization to produce value in short cycle times to meet customer needs.

The Kanban Method operates as a Pull System, that is, work is only pulled into a process step when that step has the capacity to process that work. That is, work entering that step will neither sit idle, or overload the step beyond its capacity. When step N in a workflow runs out of work (or reaches some low threshold) it signals step N-1 that it needs more work to process. Each step in a workflow has a mechanism for signaling (a Kanban) an upstream step that it is ready to accept more work. The application of WIP constraints is the central strategy for achieving this goal and to prevent the build up of queues of work in the development process. This is how teams can gain control over cycle times.

The measures of flow are Cycle Time and Throughput, and the levers of flow are WIP (Work In Process) and Work Item Age. The relationship between Cycle Time, Throughput and WIP is defined by Little’s Law:

• Cycle Time = Work in Process (WIP) / Throughput, or
• Throughput = Work in Process (WIP) / Cycle Time

Definitions

• Cycle Time is the time from when work starts on an item to when it is considered done.
• Throughput is the output volume of a production process, for example the number of user stories or bug fixes completed per unit time (day, week, sprint).
• Work in Process is the amount of accumulated work between the start and end points of a workflow.

Little’s Formula while not quite a Law of Physics, has proved to be very robust in describing the relationships between cycle time, queue (or batch) size (WIP) and processing rate (throughput). It states that the average time in a queue (cycle time) will be equal to the average queue size (WIP) divided by the average processing rate. Flow Physics.

Implications

• In a system of fixed capacity, Cycle Time is directly related to WIP. If we double the amount of work in the system, Cycle Time doubles. (Which means the cost to produce an item increases).
• WIP reduction can enable us to shorten Cycle Time without adding capacity, or changing a process.
• Reducing WIP too much reduces throughput and increases costs (idle resources).
• We can also reduce Cycle Time by reducing waste. Examples of waste include unnecessary processing steps, or processing steps that add no value to the end customer.
• We increase Throughput by increasing capacity.

Short Cycle Times are maintained by applying WIP constraints within a system. This means minimizing the accumulation of idle work items between processing steps and by preventing multitasking by overloading processing steps beyond their capacities. Cycle Time and Throughput are lagging indicators – we can only measure them once the work is done. WIP can be managed proactively in real time as the work is in process. Setting constraints on the amount of in-process work is the most straightforward and  effective way to optimize Cycle Time. We do this by limiting the admission of new work into a process step in a workflow.

Other Factors: Increasing variability of the work items leads to delays and/or buffering within the system, which will degrade the performance (Cycle Time, Throughput) of the system. This is an important consideration in software development where User Stories or ‘work items’ tend to have significant variation.

Improving flow with the Kanban Method

Applying WIP limits at each processing step in a workflow means only accepting work for processing within those limits. This is the essence of the Kanban Method. Applying Pull and WIP Limits:

• Reduces Cycle Time (per Little’s Law)
• Reduces operating expenses (time equals expense).
• Improves quality. Smaller batches and prioritizing finishing in process work over starting new work enables faster feedback loops. Problems get identified and resolved faster. Workflow policies should require that defects are fixed upon detection, and not passed downstream to the next step in the workflow. Building quality in.

The Cumulative Flow Diagram

The Cumulative Flow Diagram (CFD) visually depicts flow dynamics through a system. It is one of the most powerful tools for monitoring flow through a system, and for diagnosing problems related to queues and bottlenecks. It shows the arrival rate and departure rate of work items through a system, and shows Cycle Time (the time items are spending being processed within a system) and WIP (the total quantity of items in process at any time).

Changes in the slope of either the arrivals or departure rates indicates changes in workload demand and capacity, and their impact on WIP and Cycle Time. The CFD shows 3 of the fundamental flow metrics on a single chart.

The Kanban Method

Kanban is a method for managing Cycle Time and Throughput by visualization and WIP control. The method has 3 basic elements:

1. Defining and Visualizing a Workflow
2. Actively Managing items in a Workflow
3. Continuous Improvement of the Workflow

These elements are discussed at length here. Flow metrics are described here.

Takeaways

The underlying physics of flow enable us to see the relationships between Cycle Time, Throughput and WIP, and makes clear which levers are available to software teams for actively managing the flow of work through their development system. Visualization of workflows  exposes bottlenecks and other sources of waste, and provides direct insight into areas for improvement. “Making it Flow” means applying WIP constraints and using them to continuously manage flow and cycle time.