Improvement kata: a practice routine for reaching a condition you cannot yet see

Direction, current condition, target condition, obstacle

The sequence is what separates this from ordinary problem solving. A direction is a long-term aspiration nobody yet knows how to achieve. The current condition is measured at the process, in operating terms, by somebody who went and watched it. The target condition describes how the process should be working by a near horizon, again in operating terms. Only then does the team list obstacles and pick one. Jumping straight to the obstacle list is the usual shortcut and produces a register of irritations rather than a path, because without a target condition there is no basis for judging which obstacle matters.

A target condition describes the process, not the result

Teams write results, such as fewer defects or shorter lead time, because that is what they are asked for everywhere else. A target condition instead says how the process operates when it works: material arrives at the station in this manner, the operator sequence runs in this order without searching, the changeover begins with tooling already staged, work moves in this quantity between these two steps. Written that way it can be observed directly and argued with, progress is visible before any headline measure moves, and the team cannot reach the number by means that damage something else.

One obstacle, one experiment, one thing learned

The routine asks what you expect to happen, what actually happened and what you learned, on a short cycle. Running several changes at once feels efficient and destroys the causal information, which is the entire product of the exercise. Experiments that fail are useful and should be recorded as such. A team whose experiments always succeed is choosing safe changes it already knew would work. This is why the practice needs explicit management cover, since in most plants an engineer who tries something that does not work has some explaining to do, and that instinct has to be visibly suspended.

The coaching cycle is a claim on a manager's diary

The coach meets the learner at the process, frequently and briefly, and asks the same short set of questions: what is the target condition, what is the current condition, which obstacles are in the way, what is the next step, and when can we see what was learned. It is more demanding than it sounds, because the coach must resist answering and the meetings have to happen while the plant is busy. Attempts fail far more often on the diary than on the technique. Where senior managers do not model the routine, it becomes something the improvement team does to volunteers.

Why this is hardest for the best problem solvers

Experienced engineers and supervisors are valued for producing answers quickly, and they are right often enough for the habit to be reinforced daily. The routine asks them to withhold the answer, to let a learner run an experiment they suspect will fail, and to treat their own hypothesis as something to be tested rather than implemented. That is a genuine loss of status and speed in the short term. The compensating argument is capability, because a plant with one person who can solve problems is limited by that person's attention, which is normally the binding constraint on how fast anything improves.

Frequently asked questions

How is this different from running improvement events?

An event concentrates effort into a short burst with a defined scope and usually delivers a visible change; this is a continuing daily practice aimed at building the habit of experimenting. They combine well, since an event can establish a new condition and the routine keeps working towards the next one, but they answer different needs. Where a plant only runs events, improvement arrives in bursts and decays between them, because nobody is practising anything in the intervening weeks.

Who coaches the coaches?

Someone a level above, using the same routine, which is why the practice has to begin with senior management rather than on the shop floor. A coach who has never been coached reverts to reviewing progress and giving instructions, which looks similar and teaches nothing. Many organisations bring in an experienced practitioner for an initial period specifically to coach the coaches, then withdraw them, and that withdrawal is the real test of whether the routine took hold.

What do we do when an experiment does not work?

Record what was expected, what happened, and what that tells you about the obstacle, then choose the next step. That is the intended output, since an experiment which disconfirms an assumption has narrowed the field. The failure mode to watch for is a team quietly redefining the target condition so the result counts as a success. Keeping the target condition written down and visible, with its horizon agreed in advance, makes that hard to do without somebody noticing.