How the PDCA cycle works on the floor: four steps, three traps
In brief
Plan, do, check, act. The cycle is simple on paper; on the floor it usually stops at the second step. What the four steps mean on the floor and the three traps that leave the cycle half done.
6 November 2026 · 5 min read
Diagram: a record's life, from opening to sustainment
According to the Lean Enterprise Institute lexicon, PDCA, also known as the Deming cycle, is the four-step improvement cycle: set a goal and plan the change, make the change, compare the result with the target, and make it the standard if it worked or restart the cycle if it did not. The lexicon traces the cycle's origin to Walter Shewhart's three-step loop and to the four-step form W. Edwards Deming taught in Japan in the 1950s, and relays that the PDCA name known today was given by the Japanese Union of Scientists and Engineers in 1951. ISO 9001 also explains its process approach through this cycle. The cycle is simple; what is hard is getting as far as the fourth step on the floor.
Plan: problem and target
Planning starts not by thinking of a solution but by measuring the problem. The current state is written as a number, the target as a number, and the cause of the gap between them is sought. On the floor this step is often passed in one sentence: “Scrap is high, let's train the operator.” Yet the plan is where the cause is found and the countermeasure is tied to that cause. A plan without a found cause is a guess.
Do: on a small scale
The do step is not rolling the change out across the whole plant but trying it on one line or one shift. A small trial is cheap, reversible and teaches quickly. During the trial, what was done, when and by whom is written down; otherwise it is unclear in the third step what is being measured.
Check: measure and compare
The check step is re-measuring the number written in the plan after the trial. The measurement is done with the same method, over the same period and under the same conditions; otherwise nobody can tell whether the difference came from the change, the season or the product mix. If the result did not reach the target that is not a failure but information: the cause was guessed wrong or the countermeasure was insufficient. The date of the check is written into the plan from the start; a check without a date never happens.
Act: standard or new cycle
The fourth step forks. If the change worked it becomes the standard: the work instruction is updated, operators learn it, similar areas are offered it. If it did not work, the cycle starts with a new plan. Both paths are decisions and are recorded; the worst outcome is a trial left undecided: neither made standard nor abandoned, forgotten by itself after a while.
Three traps
Stopping at do: the most common trap. The change is made, everyone is pleased, nobody measures. Three months later the old habit returns and nobody knows why. The remedy is writing the measurement date and owner into the plan from the start.
Turning check into approval: check is the number being measured again, not a manager saying “fine”. Approval does not change the number.
Treating the cycle as one-off: PDCA is a working habit, not a project method. The change made standard is the current state of the next cycle. If the cycle turned once and stopped, what happened was a campaign, not improvement.
Small cycle, large cycle
PDCA does not turn at a single scale. The small cycle a team completes in a week and the large cycle the company turns for its annual goals follow the same logic, but not the same calendar. In the large cycle the plan is the annual goal, the doing is the projects run through the year, the check is the monthly comparison of indicators, and the act is the response to a deviation. Small cycles make up the do step of the large cycle; the check step of the large cycle says which small cycles to open. When the two scales are not linked, many improvements happen on the floor but the company's indicator does not move.
Making the cycle visible on the floor
If nobody can see which step the cycle is at, the cycle stays half done. That is why the steps of a Kaizen record follow the cycle: problem and measurement, change, result measurement, persistence follow-up. If the board shows which piece of work is waiting for a measurement, the second trap closes by itself. The manager's job is not to complete the cycle but to ask about the incomplete one: “When will we measure this?”
Sample company · demo dataKaizen record
How it is done in Gemy
PDCA is not a separate screen. The Kaizen record follows the cycle in the same order: problem and current state, the change made, result measurement and sustainment follow-up after completion. The record does not count as closed until the result is measured.
In the CAPA record the cycle runs through root cause, activity, verification, monitoring period and effectiveness check; a “not effective” decision at the effectiveness check reopens the root cause, which restarts the cycle.
Gain is kept as estimate, measurement and finance validation separately; persistence is re-measured at set intervals, and estimated and validated figures are never summed.
A change that worked is suggested to similar units through the roll-out feature; each unit gives its decision with a reason. In Hoshin the monthly Bowling chart makes the check step visible at goal level. The company switches on verified value, roll-out and Hoshin.
In short
PDCA is four steps: plan with a measured problem and target, do on a small scale, check by measuring again with the same method, act by making it standard or starting a new cycle. The three traps that leave the cycle half done on the floor are stopping at do, turning check into approval and treating the cycle as one-off. The remedy is writing the measurement date from the start and making visible which work is waiting for a measurement.
Every solved problem is a piece of knowledge; in most companies it leaves with the person who solved it. The four places where knowledge is lost, the view of ISO 30401 and ISO 9001, and five habits that make knowledge independent of people.
Takt time is the customer's pace; Yamazumi stacks each station's load against that pace. Calculating takt, reading the chart, the balancing sequence and how to explain balancing to the operator.
Autonomous maintenance means the person who knows the machine best, the operator, takes over its daily care. The content of the first three steps, the abnormality tag, the CIL standard and the division of work between operator and maintenance.
7 November 2026 · 5 min
Cookies
Essential cookies are needed for the site to work. Analytics cookies load only if you allow them; declining is as easy as accepting. You can change your choice at any time via the footer link.