Four steps, one small experiment at a time, toward a target you can see.
The improvement kata is the four-step routine at the heart of Toyota Kata. The point of the routine is to make improvement a teachable, repeatable habit rather than something only a few experienced people can do well. The steps are deliberately simple: understand the direction, grasp the current condition, set a target condition, experiment toward it. Practicing the steps daily, on real problems, develops a way of thinking that holds up under pressure. Most shops try to skip steps and end up with random improvement.
"The first instinct is always to jump to a solution. The kata is what gets you to slow down and look first."
The four steps are run in order, and the order matters.
Step one is to understand the long-term direction. This is not a goal; it is a north. Where is the organization trying to go over a three to five year horizon. Halve lead time. Win short-run work that competitors cannot quote. Reach 99 percent on-time delivery. The direction gives every short-term experiment a reason to exist.
Step two is to grasp the current condition. The learner goes to the work, watches it, measures it, and writes down what is actually happening with real numbers. Current cycle time. Current scrap rate. Current variation. The temptation to skip this step is enormous because the learner usually thinks they already know what is happening. The kata's hardest discipline is taking three or four days to study the work before forming an opinion about how to change it.
Step three is to set the target condition. This is where the learner picks a measurable target two to four weeks out. Cut average setup time from 90 minutes to 45. Reduce scrap on part number 47 from 4 percent to under 1.5 percent. The target has to be specific enough that everyone agrees when it has been hit, and close enough in time that the path to it is plausible without a complete redesign.
Step four is to experiment toward the target one small step at a time. Each experiment is a hypothesis: if we move the bin closer, setup time should drop by five minutes. The learner runs the experiment, captures what actually happened, and decides the next step based on what was learned. The cadence is short: many small experiments rather than a few big bets.
The four-step structure resembles plan-do-check-act but is stricter. PDCA can be run loosely; the improvement kata insists on the grasp step and on a specific target condition.
Imagine a 30-person machine shop where setups on the main mill take 90 minutes and the lead has been told to get them under 45. Without the kata, the lead picks two ideas, tries them both at once, can't tell which one worked, and gives up after a week. With the improvement kata, the work goes differently.
The lead spends three days videoing the next six setups and breaking them into specific tasks. That is the grasp step. They notice that of 90 minutes, 25 are tool prep that could happen before the spindle stops. The target condition is set: setups under 50 minutes within four weeks, with the tool prep moved to external time. They run one experiment per setup, change one thing at a time, and capture the result on a small log. Week one, average is 78 minutes. Week two, 64. Week four, 49. Target hit. The next target gets set: 35 minutes.
That is the improvement kata at small scale. No transformation. A discipline, run daily, by the people doing the work.
The improvement kata is one of the two routines inside Toyota Kata; its partner routine is the coaching kata, which a coach uses to develop the learner. The broader habit the kata is designed to build is kaizen, continuous small improvement by the people doing the work. The structured four-step approach is a more disciplined cousin of plan-do-check-act, the classic lean problem-solving cycle.
The questions we hear most about this term.
Long-form guides that pick up where this definition leaves off, written for manufacturers running Arda today.
Same-day setup. No distributor lock-in. Zero stockouts. Top teams double revenue in 9 months.