Toyota Kata for Continuous Improvement

At GOTO Copenhagen 2015, agile coach Håkan Forss describes using Toyota Kata to make improvement a regular learning practice. His examples come from his experience and are not controlled evidence that the method works for every team.

Forss opens with a familiar retrospective problem: teams collect many improvement ideas but act in different directions, so recurring problems return without a clear sense of progress. He calls this improvement whack-a-mole.

He explains kata as a repeated pattern that builds skill while leaving room to adapt to context. Toyota Kata includes an improvement practice and a coaching practice; the talk concentrates on improvement while emphasizing the leader’s coaching role.

The improvement pattern starts by understanding direction, observing the current condition and setting a nearer target condition. The team then runs small experiments toward that target and repeats the cycle.

A coach asks what the target is, what happened in the last experiment, what was expected and what was learned. The next obstacle, step, prediction and time to review are made explicit.

Forss recounts a prior team struggling to test software on specialized hardware. He describes using small changes and frequent conversations to make progress; this is an illustrative case rather than a controlled comparison.

He distinguishes a direction for improving the process from the business outcome itself. A distant vision provides a direction, while a target condition is a nearer hypothesis about how the work should operate.

To understand the current condition, he recommends observing work directly and collecting evidence. Visualizing intangible software work can reveal details that a high-level report misses.

He differentiates process measures, such as queue size or frequency of integration, from outcomes such as lead time, throughput and quality. The measures should be selected for the team’s goal and interpreted together.

A useful target condition should stretch the team beyond what it already knows without being impossibly distant. Take a step, observe what changes and adjust rather than assuming every obstacle can be mapped in advance.

Forss recommends recording a hypothesis and prediction before an experiment, then comparing the observed result with the expectation. His suggested failure rate is a heuristic for learning, not a universal mathematical optimum for all work.

In questions, he suggests making a proposed change small and time-limited when support is low. For leaders observing teams, he recommends asking permission, explaining intent and respecting the people doing the work.