SiliciumHex FieldKit

Problem Solving & Quality · Act

PDCA Cycle

Plan a change on a hypothesis, do it small, check the result against the prediction, then act: standardize it or start another turn.

  • Time1 h
  • FormatSmall group
  • StageAct

PDCA Cycle: what it is and why it works

PDCA, Plan, Do, Check, Act, is a learning loop for improvement. In Plan, you state the problem, the cause you believe is responsible, the change you will make and, crucially, the result you expect, in numbers. In Do, you make the change on a small scale and record what happens, including surprises. In Check, you compare the actual result with the prediction and ask what the difference teaches you. In Act, you either standardize and extend the change or adjust your hypothesis and start another cycle.

The strength of PDCA lies in its prediction. Writing down in advance what should happen turns every change into a test of your understanding, and a failed prediction becomes useful information rather than a disappointment to explain away. Small, fast cycles tend to beat large, slow projects because each one updates knowledge cheaply. Deming preferred the term PDSA, Study rather than Check, to stress learning over inspection. PDCA is the logic underneath most structured methods: the A3 report records a PDCA on one sheet, a pilot run is the Do at small scale, and standard work captures the result so the next cycle starts from a known baseline.

What you need

  • A clearly stated problem with a current value of the indicator
  • A cause hypothesis based on facts
  • A proposed change that can be tested on a small scale
  • A written, numeric prediction of the result

What you get

  • Data from the trial compared with the prediction
  • A confirmed, adjusted or rejected hypothesis
  • A standardized change or the plan for the next cycle
  • A short log of each cycle and its lesson

When to use it

When improvements are launched in one go and never checked against what was expected.

How to do it, step by step

  1. Plan: state the problem, the cause you believe in, the change you will make and the result you predict — in numbers.
  2. Do: make the change on a small scale and record what actually happened, including surprises.
  3. Check: compare the result with the prediction; ask what you learned, especially if it differs.
  4. Act: if it worked, standardize and extend; if not, adjust the hypothesis and start a new cycle.
  5. Keep a one-line log of each cycle so the learning survives the people who ran it.

Worked example: Startup scrap on a PVC pipe extrusion line

Illustrative scenario — figures are realistic but not from a real company.

A building-products plant extrudes 4-inch PVC drain pipe. Every product changeover produced about 650 lb of scrap before the pipe met wall-thickness and surface specifications. The line ran about 12 changeovers a month.

  1. Cycle 1, Plan: the team believed die temperatures were not stabilized when the line started. Change: add a 20-minute heat soak after reaching setpoint. Prediction: startup scrap below 400 lb. Do: tested on three changeovers.
  2. Check: scrap averaged 590 lb, far from the prediction. Heat soak helped slightly but was not the main cause. The operators' log showed most scrap came during the puller speed adjustment.
  3. Cycle 2, Plan: new hypothesis that operators adjusted puller speed by trial and error. Change: a startup sheet with target puller speed and screw speed for each product, derived from the best past runs. Prediction: scrap below 350 lb.
  4. Do and Check over five changeovers: average 310 lb. The prediction was met, and the result held on all shifts.

Result. Act: the startup sheet became part of standard work and the heat soak was kept only for the largest dies. Monthly startup scrap fell by about 4,000 lb. The team's log showed the first hypothesis had been wrong, and that recording it saved them from blaming temperature again months later.

Open it in the FieldKit Problem Solving app → Get the Problem Solving template pack — join the waitlist

Common pitfalls and how to avoid them

  • Skipping the prediction, so any result can be called a success.Write the expected result in numbers in Plan and compare against it in Check.
  • Implementing plant-wide in Do instead of testing small.Test on one line, one shift or a few batches first, so a wrong hypothesis is cheap.
  • Treating Check as a quick glance, then moving on.Compare data with the prediction and discuss why they differ; the gap is where learning happens.
  • Running a successful cycle but never standardizing it.In Act, update the standard, train people and set the new baseline for the next cycle.

Frequently asked questions

What is the difference between PDCA and PDSA?

PDSA replaces Check with Study. Deming preferred Study because Check can suggest simply verifying whether something worked, while Study emphasizes comparing results with the prediction and learning from the difference. The cycle is otherwise the same. In practice, many organizations use the terms interchangeably; what matters is making a prediction and learning from the result.

What is the difference between PDCA and DMAIC?

PDCA is a general, fast learning loop that can be run in hours or days on small changes. DMAIC is a structured project roadmap for complex, chronic problems where the cause is unknown and data analysis is needed. DMAIC can be seen as a detailed, data-heavy PDCA: Define and Measure are part of Plan, Analyze and Improve cover Plan, Do and Check, and Control is Act.

How long should a PDCA cycle take?

As short as the process allows. Many useful cycles take a few days, for example testing a change on several batches or shifts. Long cycles slow learning and make it harder to attribute results to the change. If a cycle needs months, consider testing a smaller part of the change or a faster leading indicator.

Origin

PDCA — Shewhart cycle (Walter A. Shewhart, 1939), taught in Japan by W. Edwards Deming from 1950; Deming later preferred PDSA.

Related methods

More in “Act”

Plan, pilot and run the fix — with owners, dates and a structured roadmap.