PDCA as a scientific method
PDCA applies the scientific method to everyday work: predict, experiment, observe, learn. A team facing a problem or opportunity forms a hypothesis ("if we change X, we expect Y"), tests it, and compares the result with the prediction. The cycle is drawn as a circle because it never ends: the learning from one turn becomes the plan for the next, which is why PDCA is the engine of continuous improvement.
For the full PDCA cycle guide, with worked examples and templates, read PDCA Cycle (Plan-Do-Check-Adjust): steps and examples.
The method sounds obvious, and that is its weakness. Many organisations run "PDCA lite": plan and do, or simply do, do, do. Changes are made on hunches, nobody checks whether they worked, and time and effort go into fixes that do not hold.
Where PDCA comes from
Walter Shewhart, the father of statistical quality control, first described a three-step cycle of specification, production and inspection, known as the Shewhart cycle. W. Edwards Deming, who edited Shewhart's work, turned it into the four-step loop he taught to Japanese engineers and managers after the Second World War, which is why it is also called the Deming cycle. Toyota built PDCA into everyday management, and it sits underneath A3 problem solving, kaizen and standard work.
PDCA vs the PDSA cycle
Deming later preferred the PDSA cycle (Plan-Do-Study-Act). "Study" stresses learning from the result rather than inspecting it. Many lean practitioners also say Plan-Do-Check-Adjust, because the fourth step is about adjusting the standard or the plan, not only acting. All three names describe the same loop and are used the same way.
The four steps
Each turn of Plan, Do, Check, Act lifts performance, and standard work stops it rolling back.
- Plan. Start with the current state: where the process is now, the objective or target, and the obstacles in the way. Break the objective into manageable steps, pick the next one, and form a hypothesis with a predicted result. Decide what data you will collect, how and when, so the check is possible. Plan also absorbs the lessons from the previous cycle's Act step.
- Do. Run the change as planned, on a small scale (one line, one shift, one product) so risk is low and feedback is fast. Do is where many teams stall in endless planning; no plan proves its value until it meets the real process.
- Check. In the check phase, compare actual results with the prediction. Did the change work, by how much, and why? Look at the process measures that produced the outcome as well as the outcome itself.
- Act (or Adjust). Reflect on what worked, what did not and what you learned. If the change worked, standardise it, train people and spread it. If not, adjust the hypothesis and start another cycle. Gather feedback from the people affected before deciding.
Questions to answer in the check phase
The check phase is the step most often skipped. Settle these questions while planning, not afterwards:
- What will be checked, and how will it be measured?
- When will we check, and who will collect the data?
- What result counts as success or failure against the prediction?
- Which process measures, not only the final outcome, will we look at?
Collecting data before and after the change on the same measure is what turns an opinion into evidence.
PDCA at different levels
PDCA applies to the macro as much as the micro. The same loop runs at every level of a well-run operation:
- Hourly and daily: a tier 1 meeting is a small PDCA cycle. The plan is the target, the do is the shift, the check is actual against target, and the act is the action assigned when there is a gap.
- Problem solving: larger problems move into an A3, which is PDCA on one page, or a kaizen event, which is one intensive cycle over a week.
- Weekly and monthly: leaders review trends, improvement work and value stream changes.
- Annually: hoshin kanri plans, executes and reviews breakthrough objectives.
These loops work best when they are connected, so that daily checks feed problem solving and problem solving feeds strategy reviews. Standard work, SOPs and tiered daily management are what hold the loops together.
PDCA vs DMAIC
PDCA is a simple, fast learning loop for everyday improving of processes. DMAIC (Define, Measure, Analyse, Improve, Control) is the Six Sigma project method: more data-intensive, slower, and suited to complex variation problems. Both follow the same logic of understanding, testing and controlling.
Common mistakes
- Spending most of the effort on Do and skipping Check, so nobody knows if the change worked.
- Planning without a clear, measurable prediction, which makes Check impossible.
- Running the change everywhere at once instead of testing small.
- Forgetting Act, so successful changes are never written into standard work and quietly drift back.
- Treating a failed experiment as a failure rather than as learning for the next cycle.
PDCA example
A maintenance team believes a new lubrication interval will cut bearing failures on a conveyor. They plan the change for one conveyor, predicting fewer unplanned stops over two months, and agree to compare stop data with the other conveyors. They do the change, check the data and find a clear improvement. They act by updating the lubrication standard for all similar conveyors, training the technicians, and scheduling a review in three months to confirm the gain holds. The review becomes the plan for the next cycle.