Service level agreements (SLAs) precisely define the responsibilities of a service provider to their customers. They can range from formal binding contracts to informal agreements. Depending on the service or industry involved, SLAs can cover service quality, availability/uptime, helpdesk hours, emergency responses, delivery times and more. Comparing service levels over time to the agreement is a good way to track performance.
Service level agreements are an efficient mechanism to keep misunderstandings, inconsistencies and disappointments at a minimum.
Service Level Agreements and Kanban
Service level agreements have typically been geared towards sectors like network services, cloud computing and outsourcing. How can SLAs be tailored to work for Kanban processes?
The core of the Kanban Method is to increase flow efficiency. Several of the flow metrics are related to this core concept, but one in particular drives customer satisfaction – cycle time. One of the fundamental practices in Kanban is that the team should pull new tasks in only when there is capacity for them. However, it is important to clarify to stakeholders how quickly tasks are likely to move through the process.
Any Kanban SLA should come with two things – a cycle time and the confidence level of achieving that same cycle time. If historically 85% of all items were completed within 24 days, the probability that any single item will be completed in 24 days or less is 85%.
Metrics reduce the guesswork involved in this process. Historical performance data is analysed to make accurate future predictions.
Estimations with Cycle Time Scatterplot
Luckily, we have a powerful tool at our disposal for estimating cycle time: the Cycle Tme Scatterplot. Every dot on the graph represents a work item from your Kanban board. The height of the dot represents its cycle time – the higher the dot, the longer it took for that task to be completed.
The dotted horizontal lines are called percentile lines. We can see that there is a 50% chance that any type of task will be completed within 6 days, and a 95% chance that it will be completed within 30 days. This means there is a 95% chance that any task, regardless of size or type, will be completed in this 30 day timeframe.
You can use Cycle Time Scatterplot percentile lines to decide service level agreements. Involve your customers and stakeholders and ask them what kind of confidence level they would be most comfortable with. We recommend starting out using the 85th percentile to define your service level agreements.
Service level agreements in Kanban are not limited to overall average cycle time. We encourage you to define SLAs with your customers and stakeholders for different types of work items or Classes of Service. For example, committing to resolving emergency tasks faster than maintenance requests. Filtering your Cycle Time Scatterplot by work item type or CoS will let you make these estimations confidently.
Estimation, Accuracy and Little’s Law
Estimation is most accurate in stable, predictable systems. The accuracy of Cycle Time Scatterplot estimations is strongly correlated to how well the process follows the Little’s Law assumptions:
- The average Arrival Rate is equal to the average Departure Rate
- All tasks entering the system will eventually exit the system once completed
- There should not be large variances in WIP between the beginning and the end of the time period examined
- The WIP average age should remain the same, neither increasing nor decreasing
- Consistent units must be used to measure Cycle Time, WIP, and Throughput
As these assumptions become less valid, the process behaviour becomes increasingly unpredictable. To keep estimations used in service level agreements as accurate as possible, the Little’s Law assumptions should be adhered to as much as possible for all different work item types (user stories, defects, maintenance requests).
Warning Signs and Taking Action
Once a service level agreement has been defined, you can use it to decide when to intervene on a lagging task. By comparing an item’s age to the agreed cycle time, we can see how the chances of missing the SLA target increase with time.
The longer an item has been in the workflow, the greater the chance that it will exceed its SLA limit. To make sure tasks are completed on schedule, you could set Kanban rules to expedite or swarm an item when it comes close to its SLA limit.
Service level agreements give you a way to assess your Kanban workflow. SLA align customer expectations and make the internal targets for your team solid but transparent. By tracking average cycle time in general and for each type of task, you can identify areas of weakness and take action to improve. When you and your stakeholders are on the same page, trust remains high and customer satisfaction is maximised.
Have you applied service level agreements to your Kanban process? Has this helped align expectations between your team, customers and stakeholders? Tell us about your experience in the comments!
