Reading the Signs: Kanban Cycle Time Scatterplot Patterns

To make your processes more efficient, first you need to identify where you are falling behind. The Kanban method uses visual methods to analyse your performance and find areas to improve. Cycle time is a key metric for quantifying how your team delivers value. Kanban cycle time scatterplot displays the cycle time for all tasks completed within a certain timeframe. Each task is represented by a dot on the chart. The height of the dot shows its cycle time.

Learning to identify the common cycle time scatterplot patterns allow you to spot problem areas before they grow bigger. Today we’ll show you how to recognise the most common scatterplot patterns and explain what they mean for your project.

1. Triangle

Meaning: A triangle with an increasing gradient means your cycle times are getting longer over time. There are two major cases to consider whenever you see this scatterplot pattern. First, tasks are coming into your process faster than they are leaving. Second, the cycle time for older items has been increased by accumulating flow debt – high priority items borrow cycle time from the rest of the tasks in order to get delivered faster.

Solution: Tighten up or lower your WIP limits. Staying vigilant about work in progress limits means arrival and departure rates can’t diverge and also keeps your team focused. No items in your process should be lying idle – this causes your team to accumulate flow debt. Lower the limit until all items in progress have at least one of your team members working on them. We recommend implementing Kanban rules to handle tasks in the order they arrive – this will stop tasks from getting stuck in the process and age unnecessarily.

2. Clusters of dots

Meaning: Clusters of dots set far apart from your usual pattern signal that something is causing unexpected drawbacks. This pattern is tricky to analyse as it can emerge in many different situations. It is a signal to start asking further questions about what is causing the clusters and consider if it is a good or bad thing.

Solution: Examine your process in detail to find the root cause of the cluster. This could be an internal or external policy, spikes in overtime, a team member out of action or any number of possible reasons. This cycle time scatterplot pattern is a warning sign to examine your process closely and ask the right questions sooner. We recommend brainstorming the potential sources of clusters with your team members.

3. Gaps

Meaning: Gaps in your scatterplot indicate that no tasks have been completed in that time period. Public holidays, vacations, strikes and government shutdowns could all be causes of gaps. This cycle time scatterplot pattern is characteristic of teams that release work in sprints rather than continuously. The equivalent pattern in the cumulative flow diagram is areas with flat lines.

Solution: Teams working in sprints will see this pattern all the time, however you should watch out for gaps that are unusually long. First, check that this pattern is not caused by any external factors (holidays etc). If you can rule these out, something is causing progress to slow down or stop completely – look closely at up and downstream processes for blockages.

4. High variability

Meaning: A highly variable Kanban cycle time scatterplot indicates a highly variable process. This is bad news for your team – this means less accurate forecasts and poor predictability.

Solution: Work on finding solutions to make your process more predictable. Are you violating the Little’s Law assumptions? Is a blockage constraining one of your steps? Are teams ignoring their WIP limits? Use Kanban meetings to brainstorm the causes of your process variability with your team and stakeholders. Be careful of tasks that just cross a percentile line – the higher the cycle time, the larger the chance of delay and of violating your service level agreements.

5. Extreme outliers

Meaning: Extreme outliers make up a sub-category of the High Variability pattern. Most of your dots are clustered predictably but a few strike out on their own. Some tasks are taking longer – much longer – to be completed than others. These are often forgotten or neglected tasks in the process.

Solution: Find and eliminate (if possible) the causes of these delays. Look out for blockages, and flag up external blockers to see if they can be resolved. To avoid tasks being forgotten or left idle, consider implementing Kanban rules to swarm tasks if they pass certain percentile lines. Extreme outliers can also be tasks that have been delayed artificially, paying the flow debt for the shorter cycle times of other work items.

Along with the cumulative flow diagram and throughput histogram, the cycle time scatterplot allows you to perform deep analysis on your process. Learning to recognise the common scatterplot patterns is a great way to identify problem areas and places to improve.

What is the most common cycle time scatterplot pattern that you observe in your process? What type of issues do you experience? How do you resolve these issues? Tell us about your experience in the comments!