Skip to content

Worked examples · 6 PDCA cycles

PDCA examples: six worked cycles from the shop floor

A PDCA cycle is one test of one change: plan it with a baseline, a target and a prediction, do it on a small scale, check the data against all three, then act by standardizing the change, adjusting it and running again, or abandoning it. Below are six complete cycles from the shop floor, one for each common kind of problem. Three met their target and their prediction. One met its target but missed its prediction, and two missed the target, which is what check is for.

Illustrative examples. These six cycles are illustrative: written from the method with realistic numbers, not taken from a real company. Every number adds up and can be recalculated from the figures given. Want the blank form? The PDCA template is free in PDF and Excel.

The six cycles at a glance

Every cycle predicts its result before the trial and checks against that prediction as well as the target. Each is the first cycle on its problem; the act step names the second.

Baseline, prediction, result and decision for each of the six PDCA examples
CycleBaselinePredictionResultAct
1. Short-length scrap on a tube sawBaseline: 3.8% scrapPrediction: about 1.1%Result: 1.1%, target metAct: Standardize
2. Staging parts for a case packer changeoverBaseline: 61 minPrediction: about 44 minResult: 50 min, target met, prediction missedAct: Adjust and run again
3. A pedestrian crossing on a forklift aisleBaseline: 15.1% don't stopPrediction: under 5%Result: 13.1%, target missedAct: Abandon this countermeasure
4. Tool search time at a setup benchBaseline: 41 min a shiftPrediction: about 8 minResult: 7.6 in week 1, 15.0 by week 3, target missedAct: Adjust and run again
5. Bore checks after a cold start on a CNC latheBaseline: 405 parts heldPrediction: 165 partsResult: 150 parts, target metAct: Standardize
6. Walking at an assembly stationBaseline: 410 m an hourPrediction: about 74 mResult: 88.7 m, target metAct: Standardize

  • Plan

    The problem with a measured baseline, the target, the suspected cause, the countermeasure, and a prediction written down before the trial.

  • Do

    Where the change was tried, what was done, for how long, and anything that went differently from the plan.

  • Check

    The result against the baseline, the prediction and the target, and what was learned from any gap.

  • Act

    Standardize, adjust and run again, or abandon this countermeasure, with the next cycle named.

Cycle 1 of 6 · Scrap

Short-length scrap on a tube saw

Where: Cut-to-length saw 2, tube cell, day shift.

Target metPrediction heldAct: Standardize

PDCA wheel for short-length scrap on a tube sawPlan: 3.8% scrap, target 1.5%, predict 1.1%. Do: Pinned stop, saw 2, day shift, 2 weeks. Check: 1.1% scrap target met. Act: Standardize on all 3 saws.PLAN3.8% scraptarget 1.5%predict 1.1%DOPinned stopsaw 2, day shift2 weeksCHECK1.1% scraptarget metACTStandardizeon all 3 sawsCycle1
Scrap on saw 2, % of tubes cut, by week
Weekly scrap ran between 3.7% and 4.0% for four weeks, then 1.2% and 1.1% in the two trial weeks, under the 1.5% target and close to the 1.1% prediction.0%1%2%3%4%5%Trial startsBeforetarget 1.5%predict 1.1%3.93.74.03.71.21.1W1W2W3W4W5W6Week

Weekly scrap ran between 3.7% and 4.0% for four weeks, then 1.2% and 1.1% in the two trial weeks, under the 1.5% target and close to the 1.1% prediction.

See the numbers
Scrap on saw 2, % of tubes cut, by week
WeekPhaseValue
W1Before3.9%
W2Before3.7%
W3Before4.0%
W4Before3.7%
W5Trial1.2%
W6Trial1.1%

Plan

Problem
Saw 2 scrapped 912 of 24,000 tubes cut in four weeks, 3.8%. The scrap tags put 641 of them (70%) down to tubes cut short, outside the plus or minus 1 mm tolerance.
Target
1.5% scrap or less on saw 2.
Suspected cause
The length stop is a sliding block held by a hand-tightened screw. Measuring 40 tubes in a row at the saw showed the length creeping shorter through a batch, and the screw had moved. Hypothesis: the stop creeps when tubes are pushed hard against it.
Countermeasure
A pinned stop: a drilled hole for each of the six lengths run on the saw and a locating pin, so the stop cannot creep, plus a go/no-go gauge for the first piece after every length change.
Prediction, written before the trial
If the stop explains the short tubes, scrap falls to what the other reasons cause today: 271 of 24,000, about 1.1%.
How it is measured
Scrap tags by reason, counted every day by the team leader, and pieces cut from the saw counter.

Do

Trial scope
Saw 2 only, day shift, all six lengths.
What was done
Maintenance drilled the stop rail and fitted the pin in three hours on a Friday. The team leader showed each operator the pin and the gauge at the start of their first trial shift.
How long
Two weeks, 12,000 tubes cut.
What went differently
On the second Tuesday a rush order was cut after a length change with the pin left out; the operator could not find it.

Check

Result
137 scrapped of 12,000: 1.1%.
Against the baseline
Down from 3.8%. Short tubes fell from about 160 a week to 4.5 a week (9 in two weeks).
Against the prediction
The prediction was about 1.1%; the result is 1.1%. The other reasons stayed where they were, as expected.
Against the target
Met: 1.1% against a target of 1.5%.
What was learned
All 9 short tubes came from the rush order cut without the pin. The pin needs a fixed home on a tether. Burrs are now the biggest reason: 78 of 137 (57%).

Act

Decision
Standardize
What changes
Pinned stops fitted to saws 1 and 3, the pin tethered to the rail, and the go/no-go check added to the saw's standard work and a one point lesson. Every shift trained before the change went live on their saw.
Next cycle
Cycle 2 on burrs, now 57% of what is left: test a fixed blade change interval on saw 2.

What this example teaches: The prediction came from the data before the trial: remove the short tubes and what is left is the other reasons. When the result lands on the prediction, you have learned that the cause was the right one, not only that scrap went down.

Templates for this kind of cycle:Defect check sheetPareto chart templateOne point lesson (OPL) template

Cycle 2 of 6 · Changeover

Staging parts for a case packer changeover

Where: Case packer, packing line 3, all shifts.

Target metPrediction missedAct: Adjust and run again

PDCA wheel for staging parts for a case packer changeoverPlan: 61 min average, target 50 min, predict 44 min. Do: Staged parts cart, line 3, 2 weeks. Check: 50 min average target met, prediction missed. Act: Adjust the list and run again.PLAN61 min averagetarget 50 minpredict 44 minDOStaged parts cartline 3, 2 weeksCHECK50 min averagetarget met,prediction missedACTAdjust the listand run againCycle1
Case packer changeovers, minutes from last good case to first good case
Eight baseline changeovers took 55 to 67 minutes (average 61). The 6 trial changeovers took 47 to 55 (average 50): at the 50-minute target, above the 44-minute prediction.3040506070Trial startsBeforetarget 50predict 44161114Changeover

Eight baseline changeovers took 55 to 67 minutes (average 61). The 6 trial changeovers took 47 to 55 (average 50): at the 50-minute target, above the 44-minute prediction.

See the numbers
Case packer changeovers, minutes from last good case to first good case
ChangeoverPhaseValue
1Before58 min
2Before64 min
3Before55 min
4Before61 min
5Before67 min
6Before59 min
7Before61 min
8Before63 min
9Trial52 min
10Trial49 min
11Trial47 min
12Trial55 min
13Trial48 min
14Trial49 min

Plan

Problem
Format changes on the case packer take 61 minutes on average, last good case to first good case, over the last 8 changeovers (55 to 67). On two filmed changeovers the crew spent 16 and 18 minutes of the stop walking to the store for change parts and tools.
Target
50 minutes or less, so two changeovers a shift fit in the planned changeover time.
Suspected cause
Change parts and tools are fetched after the line stops because nothing tells anyone to fetch them earlier. Hypothesis: the fetching is external work done as internal work.
Countermeasure
A changeover cart, loaded by the team leader in the 30 minutes before the stop from a parts list for each case size.
Prediction, written before the trial
The stop falls by the fetching time: 61 minus 17 gives about 44 minutes.
How it is measured
Changeover log from last good case to first good case, and two trial changeovers filmed.

Do

Trial scope
Line 3 case packer only, every format change for two weeks.
What was done
The maintenance planner and the team leader wrote the parts list for each case size from the films. A spare cart was painted and labelled.
How long
Two weeks, 6 changeovers.
What went differently
In 4 of the 6 changeovers someone still left the line for a part: the glue nozzle set and the side guides for the small case are kept in the maintenance crib, and they were not on the list.

Check

Result
50 minutes on average (47 to 55).
Against the baseline
11 minutes shorter than the 61-minute baseline. No trial changeover took longer than the quickest baseline one (55 minutes).
Against the prediction
Missed: 50 minutes against a prediction of 44. Fetching on the filmed trial changeovers still took 5 and 7 minutes, so the cart removed 11 of the 17 minutes, which is the whole gain.
Against the target
Met, just: 50 minutes against a target of 50 or less.
What was learned
The theory held (fetching was the waste to move) but the parts list was incomplete. Hitting the target hid that; the gap to the prediction showed it.

Act

Decision
Adjust and run again
What changes
A second glue nozzle set and the small-case side guides now live on the cart, and the list is checked against one filmed changeover per case size. Same line, same measure, two more weeks.
Next cycle
Cycle 2 predicts about 44 minutes again. After that, the trial cases and adjustments (19 and 20 minutes on the films) are the biggest block left: mark the guide settings for each size.

What this example teaches: Meeting the target is not the same as understanding the result. Because a prediction was written down, the six-minute gap was visible and pointed straight at the incomplete parts list.

Templates for this kind of cycle:Changeover checklistSMED changeover worksheet

Cycle 3 of 6 · Safety near miss

A pedestrian crossing on a forklift aisle

Where: Pedestrian walkway crossing the main forklift aisle by the shipping dock.

Target missedPrediction missedAct: Abandon this countermeasure

PDCA wheel for a pedestrian crossing on a forklift aislePlan: 15.1% don't stop, target under 5%. Do: Mirror and STOP line, 1 crossing, 2 weeks. Check: 13.1% don't stop target missed. Act: Abandon; next: gate and guardrail.PLAN15.1% don't stoptarget under 5%DOMirror and STOP line1 crossing, 2 weeksCHECK13.1% don't stoptarget missedACTAbandon; next:gate and guardrailCycle1
Crossings without a stop to look, % of crossings observed
  • Before: Before (47 of 312)15.1%
  • Prediction: Predicted and targetunder 5%
  • Trial: Trial (39 of 298)13.1%

Conflicts (someone had to brake or step back): 6 before, 5 in the trial, in 5 hours observed each time.

Before: 15.1% of 312 crossings. Predicted: under 5%. Trial: 13.1% of 298, far above the target.

Plan

Problem
Five near misses between a forklift and a pedestrian were reported at this crossing in three months. Too few to measure a two-week trial, so the team measured behaviour: in 10 observations of 30 minutes at shift peaks, 47 of 312 people (15.1%) stepped into the aisle without stopping to look, and 6 times a truck or a person had to brake or step back.
Target
Under 5% of crossings without a stop, and no conflicts.
Suspected cause
Racking on the dock side hides trucks until they are about three metres from the crossing. Hypothesis: people do not stop because they cannot see what is coming, and nothing reminds them.
Countermeasure
A convex mirror on the rack end and a painted STOP and LOOK line on the walkway at the aisle edge.
Prediction, written before the trial
Unsafe crossings fall below 5%.
How it is measured
The same 10 observations at the same times, by the same two observers, counting each crossing as stopped or not and each conflict.

Do

Trial scope
One crossing.
What was done
The mirror went up and the line was painted on a Sunday. The supervisor announced the change at the start-of-shift meetings.
How long
Two weeks, with the observations spread over both weeks so the first days, when the change was new, did not dominate the count.
What went differently
People who did stop used the mirror. Most of the people who did not stop were carrying something or walking in pairs.

Check

Result
39 of 298 crossings without a stop (13.1%), 5 conflicts.
Against the baseline
From 15.1% to 13.1%, and conflicts from 6 to 5: a difference too small to tell apart from chance with this many observations.
Against the prediction
Missed by a wide margin: the prediction was under 5%.
Against the target
Missed: 13.1% against under 5%, and conflicts still happen.
What was learned
Seeing the truck was not the main problem; stopping was. A sign and a painted line ask people to choose to stop every time, and busy people do not. NIOSH ranks controls that depend on people's ongoing effort below engineering controls that block the hazard.

Act

Decision
Abandon this countermeasure
What changes
The painted line and the sign are dropped as the countermeasure. The mirror stays, because it helps the people who do stop, but no one counts it as the fix.
Next cycle
Cycle 2 tests an engineering control: a guardrail along the walkway with a self-closing gate set back from the aisle edge, so a person has to stop and open it before stepping out. Prediction: under 2% of crossings without a stop.

What this example teaches: A failed check is a result, not a failure of PDCA. Two weeks and a tin of paint bought the knowledge that this crossing needs a physical stop, before anyone was hurt. Measuring behaviour made a two-week test possible where near-miss reports were too rare to show anything.

Templates for this kind of cycle:Near miss report formSafety observation card templateFloor marking color guide

Cycle 5 of 6 · Inspection frequency

Bore checks after a cold start on a CNC lathe

Where: CNC lathe 4, one part number, bore diameter checked with a bore gauge.

Target metPrediction heldAct: Standardize

PDCA wheel for bore checks after a cold start on a cnc lathePlan: 405 parts held, target 202 or fewer, predict 165. Do: Checks every 20 min, in warm-up, 4 weeks. Check: 150 parts held target met. Act: Standardize in the control plan.PLAN405 parts heldtarget 202 or fewerpredict 165DOChecks every 20 minin warm-up, 4 weeksCHECK150 parts heldtarget metACTStandardize inthe control planCycle1
Parts held for sorting in four weeks, lathe 4
  • Before: Before: hourly checks405
  • Target: Target202 or fewer
  • Prediction: Predicted165
  • Trial: Trial: every 20 min in warm-up150

Checks a shift: 8 before, 12 in the trial (8 more minutes of checking a shift).

Hourly checks held 405 parts in four weeks. The prediction for the new check plan was 165 and the target 202 or fewer; the trial held 150.

Plan

Problem
The operator checks the bore every hour, 8 checks a shift. In four weeks (20 shifts, 160 checks) 9 checks failed, and 8 of the 9 were in the first two hours after a cold start. Each failed check puts every part made since the last good check on hold for sorting: with hourly checks, an hour of output, 45 parts.
Target
Half the parts held or fewer: 202 or fewer in four weeks.
Suspected cause
The bore drifts while the machine warms up, then settles. Hypothesis: the checks are spread evenly over a shift, but the risk is not; they are too far apart in the first two hours.
Countermeasure
Check every 20 minutes for the first 2 hours after a cold start, then hourly: 12 checks a shift instead of 8.
Prediction, written before the trial
If failures happen as before (8 early, 1 later), each early one holds 15 parts instead of 45: 8 times 15 plus 45 gives 165 parts held in four weeks.
How it is measured
The check record (time, reading, pass or fail) and the hold tags.

Do

Trial scope
Lathe 4 and its one part number only.
What was done
The quality engineer changed the check sheet for lathe 4, with the early check times printed on it. The operators agreed the cost: 4 more checks a shift at about 2 minutes each, 8 minutes a shift.
How long
Four weeks, 20 shifts, 240 checks.
What went differently
Two shifts started after a long stop for a breakdown; the operators asked whether that counts as a cold start. It was treated as one.

Check

Result
8 failed checks (7 early, 1 later) and 150 parts held.
Against the baseline
150 parts held against 405: 63% fewer to sort.
Against the prediction
Close: 150 against a predicted 165, with one fewer early failure than the baseline weeks.
Against the target
Met: 150 against 202 or fewer.
What was learned
The drift still happens: 7 of 8 failures were early again. Checking more often catches it sooner but does not stop it.

Act

Decision
Standardize
What changes
The new check frequency goes into the control plan for this part on lathe 4, with the rule the operators asked for written in: a stop of more than 30 minutes counts as a cold start.
Next cycle
Cycle 2 goes after the cause: a 15-minute warm-up routine before the first part. Prediction: early failures fall to one or none in four weeks. If that holds, the extra early checks can go.

What this example teaches: Inspection frequency is a trade: more checks cost minutes, fewer checks cost sorting when one fails. Put the checks where the risk is, and treat the result as containment while the next cycle works on the cause.

Templates for this kind of cycle:Control plan templateI-MR chart template (individuals and moving range)

Cycle 6 of 6 · Walking distance

Walking at an assembly station

Where: Assembly station 3, final assembly line, both shifts.

Target metPrediction heldAct: Standardize

PDCA wheel for walking at an assembly stationPlan: 410 m an hour, target under 100 m, predict 74 m. Do: Flow rack, printer, station 3, 1 week. Check: 88.7 m an hour target met. Act: Standardize, copy to 2 and 4.PLAN410 m an hourtarget under 100 mpredict 74 mDOFlow rack, printerstation 3, 1 weekCHECK88.7 m an hourtarget metACTStandardize,copy to 2 and 4Cycle1
Metres walked an hour at station 3, by destination
  • Before: Before410 mFastener rack 264 m, Label printer 90 m, Other 56 m
  • Prediction: Predicted74 mLabel printer 18 m, Other 56 m
  • Trial: Trial88.7 mFastener rack 14.7 m, Label printer 18 m, Other 56 m

Target: under 100 metres an hour.

Before: 410 metres an hour, 264 of them to the fastener rack and 90 to the printer. Predicted: 74. Trial: 88.7, under the 100-metre target.

Plan

Problem
Traced on a spaghetti diagram for three hours, the station 3 operator walked 412, 388, 430 metres: 410 metres an hour, about 3,075 metres in a 7.5-hour shift. 264 metres an hour were trips to the central rack for fasteners (6 round trips of 44 metres) and 90 to the label printer (3 of 30 metres).
Target
Under 100 metres an hour.
Suspected cause
The two fastener sizes used on every unit are stored with the rest in the central rack, and the label printer sits where the old line layout put it. The walking is built into the layout, not the operator's method.
Countermeasure
A two-lane flow rack at the station for the two fastener sizes, refilled by the material handler on the hourly route, and the label printer moved to the station.
Prediction, written before the trial
Rack trips stop and the printer trip shrinks to 6 metres: 18 plus 56 other gives about 74 metres an hour.
How it is measured
Three more traced hours with the same measuring wheel, during the trial week.

Do

Trial scope
Station 3 only, both shifts.
What was done
A quick kaizen: the flow rack and the printer were moved in a lunch break, and an electrician added an outlet. The material handler's route sheet got one more stop.
How long
One week.
What went differently
In the first traced hour the operator built one unit of a variant that uses a third fastener size, still kept in the central rack.

Check

Result
118, 70, 78 metres in the three traced hours: 88.7 metres an hour.
Against the baseline
From 410 to 88.7 metres an hour; about 665 metres a shift instead of 3,075.
Against the prediction
Above the 74-metre prediction by the one 44-metre trip for the third fastener size. Without it the three hours average 74 metres, exactly as predicted.
Against the target
Met: 88.7 against under 100.
What was learned
The change works as planned. The one variant shows that a point-of-use list has to cover every variant built at the station, not only the common ones.

Act

Decision
Standardize
What changes
The flow rack and printer positions are drawn into the station layout and its standard work, the refill is on the material handler's route, and a small bin for the third fastener size is added to the rack.
Next cycle
Copy it to stations 2 and 4, which have the same layout (yokoten). Then re-time station 3: walking saved is not output gained until the station's cycle time shows it.

What this example teaches: A quick kaizen still gets a full cycle. A measured baseline, a number to beat and a second set of traces turn a lunch-break move into a standard that can be copied to the next station with confidence.

Templates for this kind of cycle:Spaghetti diagram templateQuick kaizen templateWater spider route plan (mizusumashi)

PDCA or PDSA? Where the cycle comes from

The cycle is older than lean and has been renamed several times. Dates and steps below follow Moen and Norman's history of the cycle and the Lean Lexicon.

  1. 1939

    Shewhart's cycle

    Walter Shewhart, in Statistical Method from the Viewpoint of Quality Control, turned specification, production and inspection from a straight line into a circle, and compared the three steps to making a hypothesis, running an experiment and testing the hypothesis.

  2. 1950

    The Deming wheel

    At a seminar for the Union of Japanese Scientists and Engineers (JUSE), W. Edwards Deming presented a four-step version: design the product, make it, sell it, then test it in service and find out what users think, and go round again.

  3. 1951

    Japanese PDCA

    Japanese executives recast the Deming wheel as plan, do, check, act, according to Masaaki Imai; no one person claims authorship. The Lean Lexicon dates the change to JUSE in 1951. Act meant standardizing what worked or going back to plan, and Kaoru Ishikawa later added goals and methods to plan and training to do.

  4. 1986

    Deming's Shewhart cycle

    In Out of the Crisis, Deming taught a version in which the change or test is carried out, preferably on a small scale, and the results are studied. In his seminars he warned that check is the wrong word.

  5. 1993

    PDSA

    In The New Economics, Deming called it the Shewhart cycle for learning and improvement: plan, do, study, act. Act means adopt the change, abandon it, or run through the cycle again.

  6. 1991 and 1994

    Prediction and the Model for Improvement

    Moen, Nolan and Provost made a prediction part of the plan and had the study step compare the data with it. Langley, Nolan and Nolan added three questions in front of the cycle: what are we trying to accomplish, how will we know a change is an improvement, and what change can we make.

Deming did not like PDCA. Writing to Ronald Moen in 1990 about a manuscript, he asked for the cycle to be called PDSA, "not the corruption PDCA". The Deming Institute puts the difference this way: check looks at whether a plan succeeded or failed, while study compares the result with a prediction, to learn and revise the theory behind the change.

Moen and Norman conclude that the two cycles are related only through the scientific method: PDCA came out of Japanese quality control and is mostly used to implement and standardize, PDSA is mostly used to test and learn.

This page says PDCA because that is what most plants call it. But every example writes a prediction before the trial and studies the gap afterwards, which is the PDSA habit. The changeover cycle shows why it matters: it met its target, and only the missed prediction showed that the parts list was incomplete.

Test on a small scale first

Deming's 1986 and 1993 versions of the cycle both say to carry out the change preferably on a small scale. Every example above does: one machine, one crossing, one bench or one station. Four rules for sizing the trial:

  • Small enough to be cheap to undo

    One saw, one crossing, one bench, one station. If the change is wrong, it is wrong in one place for two weeks, not in the whole plant.

  • Large enough to cover the conditions that matter

    The 5S cycle tested on the day shift only, and the night shift undid it. IHI's guidance is to test under varying conditions (shifts, days, products) before calling a change ready.

  • Long enough to see a trend, short enough to learn fast

    Two to four weeks was enough for every example here. Set the check date in the plan, and plot each day or week, so a slip like the one at the setup bench shows up.

  • Measured the same way before and after

    Same counter, same observers, same hours of the day. If the measure changes with the change, you cannot tell which one moved the number.

Write your own cycle: a checklist

  1. Measure the baseline before you change anything, the same way you will measure the result.
  2. Write a target and a separate prediction: the target is what you need, the prediction is what you expect this change to do and why.
  3. Name the cause you are testing. If the countermeasure does not act on that cause, the cycle cannot teach you anything.
  4. Pick a trial scope that is small but includes every shift, product or person that touches the process.
  5. When a measure is too rare to move in weeks (injuries, near misses), measure the behaviour or condition that leads to it.
  6. Compare the result with the prediction as well as the target. A met target with a missed prediction still has something to teach.
  7. End with a decision: standardize (update the standard work and train every shift), adjust and run again, or abandon. Then write down the next cycle.

Write it on the PDCA template. When the problem needs more room for the cause, use an A3 and find the cause with the 5 Whys.

Sources

The history and the method follow these sources. The six cycles, their numbers and the notes are ours.

Free templates and tools for your own cycle

The blank PDCA and A3 forms, a root cause sheet, and calculators for the before and after numbers.

Go deeper

Key terms

Lean glossary

Doing this in LeanSuite

FAQ

Questions about PDCA examples

More examples

See all example pages and the lean glossary.

Running PDCA cycles on paper and email?

See on a call how LeanSuite runs PDCA, A3 and DMAIC projects from built-in templates and tracks projected and actual savings for each one.

Schedule Demo