Improvement Kata storyboard

The Improvement Kata is a practice routine for reaching a goal when you can't see the path in advance. ("Kata" means a pattern you practise until it becomes habit.) Mike Rother described it in his book Toyota Kata, from studying how Toyota develops improvement thinking in its people. It has four steps: understand the challenge, grasp the current condition, set the next target condition, and experiment toward it one step at a time. The storyboard is the one sheet the learner keeps for all four, and the sheet a coach reads with them in a short, regular coaching cycle. It deliberately doesn't hold a plan of solutions. The route is found through small experiments, each with a prediction written down before the step, so the gap between what you expected and what happened shows exactly where your knowledge ran out, which is where the next step goes.

Challenge

Write the direction you're working toward, usually six months to three years out, set by or agreed with the level above you. It should be beyond what you know how to do today: if you already know how to get there, it's a project plan, not a challenge. Describe what will be different for the customer or the process, with a measure and a date, and say why it matters. Don't write the solution into the challenge.

Current condition

Go and see the work where it happens before setting a target condition. Watch several cycles, time them, and talk to the people doing the work. Write down two things. The operating pattern: how the work actually runs today, including the sequence, who does what, where time goes and how much it varies. And the outcome: what the process produces today, with its spread, not just an average. Date each observation, and note where you were, what you saw and who does the work. Reports and dashboards can show you where to look, but they don't replace watching. Update the current condition after each step; it keeps moving.

Next target condition

Describe how you want the process to be running on a date ahead. Anything from about a week to about three months works; 1 to 4 weeks is a good default, because short cycles mean you learn faster. Go longer only when the process itself is slow, for example when each result takes weeks to show. Write it in two parts. First the operating pattern, as a state you could go and see: "every changeover runs from materials already staged at the booth, and both shifts clean in the same order." Then the outcome metric, with a number, a unit and a date: "35 minutes or less on 8 of the last 10 changeovers, by 15 May." It isn't a list of actions: "train the operators, buy a cart" is a to-do list, and the actions are what your experiments will discover. It should sit beyond what you know how to reach today. Once it's set, keep it until the date, and put what you learn into the next one.

Obstacles parking lot

List what you think stands between the current condition and the target condition. It's a parking lot, not a plan: you don't need to rank it carefully or solve everything on it. Pick one obstacle, mark it in focus, and work only on that one until it's cleared or you learn it doesn't matter. Expect the list to change as you learn; add new obstacles as they appear and strike out cleared ones. Obstacles are about the process (the method, materials, equipment or information), never about a person.

Experiment records

One row per step, filled in as a Plan-Do-Check-Act (PDCA) cycle. Plan: write the step and what you expect will happen, before you take it. Do: take the step. Check: write what actually happened, as soon as you can go and see. Act: write what you learned, which shapes the next step. There are two kinds of step. An observe step goes and looks to learn more, and changes nothing. A change step changes one thing, so you can tell what caused the result. Give each step a date to go and see the result, ideally within a day or a few days; if you can't see the result within about a week, the step is too big, so break it down. Never fill in or edit the expectation once you know the result. A prediction that didn't come true is the most useful row on the sheet.

Threshold of knowledge

The threshold of knowledge is the line between what you know and what you don't know yet. Before each step, write what you don't know yet that the step should find out. Your next step belongs right at that line: not so far beyond it that you can't interpret the result, and not behind it repeating what you already know. When a result surprises you, you've found the line. Not knowing is normal and useful; pretending to know is what stalls improvement.

Metric run chart

Record the outcome metric on the day you measure it, from your own measurements, and plot it as a run chart: dates along the bottom, values up the side, with the target line drawn across. Look at the pattern over several points rather than reacting to a single one. Don't rebuild the numbers afterwards from memory.

Coaching cycle

A short, regular check-in at the storyboard: daily if you can, a few times a week at least, and short enough to fit between other work. The learner walks the coach through the board from left to right in their own words: where they're heading and the target they're working toward now, what they saw the last time they watched the work, what they expected from their last step and how it turned out, which obstacle has their attention, and what they'll try next and what they predict. The coach listens for gaps in the reasoning (a prediction written after the result, a claim nobody went and looked at, a step that changes several things at once) and asks how the learner knows. The coach doesn't hand over answers; together they agree when to look at the result of the next step. The routine draws on the coaching practice described in Mike Rother's Toyota Kata; the wording here is Katafacts' own.

The example below continues Line 2's paint booth after the sequencing rule. The challenge is changeovers under 10 minutes by the end of 2027, so custom colours no longer wait for set slots. The current condition comes from watching changeovers on both shifts, with one report entry clearly marked as a report. The target condition, three weeks out, describes how changeovers should run plus "35 minutes or less on 8 of the last 10". There are four experiment records: an observe step that timed the walks away from the booth, two single-change steps (a staged cart, then the nozzle size on the queue card), and a new observe step on the obstacle now in focus, still waiting for its result. Two obstacles are cleared, one is in focus, and a run chart shows changeover time falling toward the target line.

• The target condition is a to-do list.

  "Buy a cart, train second shift, update the standard" decides the answers before anyone has learned anything, and leaves nothing to check except whether the tasks got done. Describe how the process will be running, plus a dated outcome, and let the experiments find the actions.

• The prediction is written after the result.

  Once you know what happened, it's almost impossible not to "expect" it. The gap between expected and actual is what shows where your knowledge ran out, so write the expectation first and never edit it.

• Several changes in one step.

  If you stage a cart, move the solvent store and change the cleaning order all at once, you can't tell which one made the difference, or which one made things worse. Change one thing per step. Observe steps, which change nothing, are how you learn before changing.

• Skipping go and see.

  A current condition written from a report or from memory describes the process as people think it runs. The walks away from the booth, the drying wait and the locked cage only show up when you watch. Go again after every few steps, because the current condition keeps moving.

• Blaming people.

  "Operator error", "retrain the team" and "be more careful" stop the learning at a person and leave the process that let it happen untouched. Ask what about the method, materials, equipment or information made the mistake possible, and work on that.

• Steps too big to learn from.

  A step whose result you can't see for weeks gives you a few learning cycles a quarter instead of several a week. If the go-and-see date is more than about a week away, break the step down.