By Travis Birch
So you are embarking on an Agile transformation. One important thing you need to figure out (and hopefully you have some consultants helping you) is how many Scrum Teams you will need for product development and how many Kanban teams you will need for operations, deployment, support, maintenance, keeping the lights on, etc. Because when you create a bunch of cross-functional teams they will have gaps in their Agile engineering practices and their definition of “done” will have to be limited by several organizational factors. The teams won’t have access to many of the environments and there won’t be enough specialized resources to assign to each team. Plus, the nature of the work coming into the ops and support teams are much more finely grained and varied than user stories in a Sprint Backlog and it’s important that the Scrum teams focus on their products and are not disrupted by every little change request from the business. So you need Kanban teams to do the work that the Scrum Teams can’t do yet as well as the stuff that you don’t want your Scrum Teams to be distracted by.
That’s more or less the popular consensus on the role of Kanban in the Agile community. Some acclaimed thought leaders have even written popular books and blog posts about it. But if one digs a little deeper, one finds that there is much more to Kanban than meets the common agilist’s eye (although clearly this is changing, evidenced by the existence of this piece of writing). So here are the 14 things that I can think of off the top of my head:
- The Kanban Method is not about transformation. At least not the radical, deep Satir J-Curve brand of transformation that has been attempted in many organizations. All too often with the radical approach, there is enough resistance to change and enough ensuing chaos for the leaders of the organization to lose patience with the change initiative before the system has a chance to recover.
- Moreover (and dangerously), many so-called champions of change are not aware that Satir was a psychotherapist and that her J-Curve method was employed to change the identity of her clients, breaking them down and building them up again. What’s important here is that Satir was a psychology professional working with willing clients to help them transform into the people they wanted to be. This is not the case with most professional knowledge workers. Knowledge workers tend to not be seeking this kind of service from managers and coaches. A more brutal version of this technique has been employed to transform teenagers into deadly soldiers.
- Instead of deep J-Curve transformation, the Kanban Method proposes small, rapid J-Curve experiments. This approach to change provokes less resistance, avoids extended periods of thrashing in chaos and the severe testing of leadership patience. Well-designed small J-Curve experiments produce just enough organizational stress to stimulate change without drop-kicking people into deep chaos and despair and forcing the regression of an organization to lower levels of trust and maturity.
- The Kanban Method is a management system for the design and evolution of the interdependent services of an organization. Such services are often composed of several Scrum Teams, shared services, managers, senior staff and specialists and are often themselves served by other services. Every service is delivered via a system of stages of knowledge discovery (rather than hand-offs). See more on this here.
- The design of a system determines the fitness for purpose, flow of value and quality of the service (as demonstrated by Deming’s Red Bead Experiment). It’s not about high-performance teams. It’s about the performance of the system.
- Transparency of the system empowers knowledge workers to self-organize around the work because they understand the system and are trusted to know what to do in order to deliver the service.
- Kanban managers are systems managers, not people managers and not coaches.
- Team-level Kanban is actually a form of proto-Kanban—still Kanban, but in an immature state, an incomplete rendering of a service delivery system.
- A Kanban system is a pull system. The capacity of the system is calibrated for optimum flow. New work enters the system when there is sufficient capacity to absorb the new work without overburdening the system and disrupting flow. Demand is balanced against capacity.
- All demand is potentially refutable. When there is capacity in the system to start new work, sources of demand collaborate to determine what is the most important work to start next and the system is replenished.
- Deciding what to start next is based on economics—transparent and rational risk assessment.
- Once the system is replenished and there is a commitment to deliver the newly-started work, risks are managed with explicit policies such as classes of service, work-in-process limits, pull-readiness criteria, feedback loops and relevant metrics (i.e. not team velocity).
- Average lead time from project or feature commitment to completion is a basic metric. Improving the system results in a reduction in both lead time and lead time variability. Delivery forecasts are based on historical lead time data. Deadlines are also managed with lead time data (i.e. deciding when to start something).
- All of the above is the responsibility of management. This should leave little management capacity for monitoring individual performance and story point velocity of teams (white bead count). A sign of a mature Kanban system is that managers have improved their behaviour and are focused on improving the system and that knowledge workers are free to self organize around the work as skilled, adult professionals.