The PDCA cycle (Plan-Do-Check-Act) is the backbone of any modern continuous improvement system. Popularised by W. Edwards Deming from Walter Shewhart's cycle in the 1940s, this iterative four-phase method remains, eighty years later, the operational framework that the ISO 9001:2015 standard implicitly requires in its clause 10.3 on continual improvement. In this article we develop each phase with technical depth, show how to apply it on a real manufacturing line, and point out the mistakes that most often cause its implementation to fail.
What the PDCA cycle is and why it is still relevant
PDCA is a closed loop of organisational learning. Its value lies not in any of its phases in isolation, but in the disciplined repetition of the complete loop: each turn converts the empirical knowledge of the previous turn into a new working standard. Deming insisted that the cycle never truly "closes"; on completing the Act phase, the improved standard becomes the new starting point for the next Plan phase. This is the essential difference between a one-off improvement (a project that ends) and continuous improvement (a process that never ends).
The ISO 9001:2015 standard structures its entire Annex SL around this logic. Clauses 4 to 7 correspond to Plan, clause 8 (Operation) to Do, clause 9 (Performance evaluation) to Check, and clause 10 (Improvement) to Act. Whoever understands PDCA understands the whole architecture of the standard. It is also worth distinguishing PDCA from its variant PDSA (Plan-Do-Study-Act), which Deming himself preferred in his later years because "Study" conveys the idea of reflective analysis, rather than mere inspection, better than "Check".
Plan phase: planning with testable hypotheses
Planning is not drawing up a to-do list. It is formulating a falsifiable hypothesis: "we believe that if we reduce the extrusion temperature from 230 °C to 215 °C, the bubble-rejection rate will fall from 3.2% to 1.5%." A good Plan phase requires four elements. First, an operational definition of the problem supported by data, not opinions. Second, a root-cause analysis using tools such as the Ishikawa (fishbone) diagram or the 5 Whys. Third, a quantified objective with its metric and its baseline. Fourth, an action plan with owners, deadlines and resources.
Kaoru Ishikawa's seven basic quality tools are applied almost entirely in this phase: data collection sheets, histograms, Pareto charts to prioritise the vital causes over the trivial ones, scatter diagrams to detect correlations, and stratification to segment the data by shift, machine or batch. The Pareto chart usually reveals that 80% of defects come from 20% of the causes, which allows effort to be concentrated where it pays off most.
Do phase: executing on a small scale
The most expensive mistake in PDCA is jumping from Plan to a full-scale rollout. The Do phase must first be run as a controlled pilot trial: one shift, one machine, a limited batch. The aim is not to solve the problem yet, but to generate clean data on whether the hypothesis holds. During the pilot everything is recorded: process conditions, measurements, incidents and deviations from the original plan. That traceability is what later allows verification to be carried out with rigour.
In manufacturing, the Do phase is best supported by standardised work: a standard work sheet documenting the sequence, the cycle time and the work-in-process inventory. Without a prior standard it is impossible to know whether what changed was the proposed improvement or simply each operator's way of working. Toyota summed it up in a maxim: "without a standard there is no improvement, only change."
Check phase: verifying against the hypothesis, not against the wish
Checking means comparing the actual results of the pilot with the prediction made in the Plan phase. This is where Statistical Process Control (SPC) comes in: X-bar and R control charts for continuous variables, p or np charts for attributes. A process is considered to be under statistical control when it shows only common-cause variation; signals beyond the control limits (±3 sigma) indicate special causes that must be investigated. The Cpk capability index measures whether the process, in addition to being under control, meets the specifications: a Cpk ≥ 1.33 is the usual acceptance threshold in the automotive sector, and ≥ 1.67 in critical sectors.
Analytical honesty is decisive. If the data refute the hypothesis, that is not a failure of PDCA: it is exactly what the cycle is meant to detect before investing in a full-scale rollout. Verifying against what was wished rather than against what was predicted is the most subtle way of corrupting the method.
Act phase: standardise or adjust
The Act phase branches according to the result of Check. If the improvement is confirmed, it is standardised: the work instruction is updated, staff are trained, the control plan is modified and it is rolled out to all lines. If the improvement is not confirmed, or only partly so, the hypothesis is adjusted and a new Plan cycle begins with the lessons incorporated. In both cases, what has been learned is documented so that it is not lost as staff turn over.
Standardisation is what distinguishes PDCA from mere trial and error. Without it, each improvement evaporates as soon as the shift changes. That is why many plants combine PDCA with SDCA (Standardize-Do-Check-Act): while PDCA raises the level, SDCA maintains it, preventing the process from "slipping" backwards between improvement cycles.
Worked example: reducing rejects in plastic injection moulding
An injection moulding plant detects a 3.2% reject rate due to flash. Plan: the Pareto chart shows that 78% of the flash is concentrated in one four-cavity mould; the 5 Whys lead to insufficient clamping pressure caused by plate wear. Hypothesis: increasing the clamping force from 80 to 95 tonnes will bring the defect below 1%. Do: it is trialled over two shifts on a single machine, recording 1,200 parts. Check: the p chart shows that rejects fall to 0.8% and the process comes under control; Cpk rises from 0.9 to 1.4. Act: the process sheet is updated, all three shifts are trained, and preventive plate replacement is scheduled for the remaining moulds. The new standard becomes the baseline for the next cycle.
Comparison table: PDCA versus other improvement frameworks
| Framework | Focus | When to use it | Standard or reference |
|---|---|---|---|
| PDCA | Iterative incremental improvement | Chronic problems and day-to-day improvement | ISO 9001:2015 cl. 10.3 |
| DMAIC (Six Sigma) | Statistics-based variation reduction | Complex problems with abundant data | ISO 13053 |
| Kaizen | Rapid improvements in short events | Intensive 3-5 day workshops | Lean philosophy |
| A3 (Toyota) | Problem-solving on a single sheet | Communicating and solving at the same time | Toyota Production System |
Common mistakes that cause PDCA to fail
- Skipping the Check: rolling out the improvement without verifying the pilot, assuming it will work.
- Plans without a baseline: with no starting figure, no improvement can be demonstrated.
- Not standardising in Act: the improvement is lost when the shift or operator changes.
- Confusing activity with results: measuring how many actions were taken instead of their effect on the metric.
- Cycles that are too long: a PDCA that lasts six months loses the rhythm of learning; many short cycles are preferable.
Frequently asked questions
How many turns of the PDCA cycle are needed?
There is no fixed number. Continuous improvement means that the cycle never ends: each standard reached is the starting point for the next one. In practice, it is preferable to complete many short cycles (weeks) than a few long ones.
Is PDCA mandatory for ISO 9001 certification?
The ISO 9001:2015 standard does not literally mention "PDCA" as a requirement, but its entire structure is based on this cycle and clause 10.3 requires continual improvement. A management system without an iterative improvement mechanism will struggle to pass the certification audit.
What is the difference between PDCA and Kaizen?
Kaizen is the philosophy of continuous improvement through small steps; PDCA is the concrete operational method that structures each of those improvements. Kaizen is the cultural "what" and PDCA is the technical "how".
Is PDCA useful outside manufacturing?
Yes. Although it was born in industry, the cycle is applied successfully in services, healthcare, education and administrative processes. The key is always the same: formulate testable hypotheses, trial them on a small scale and standardise what works.
Conclusion
PDCA is not just another tool in the quality catalogue: it is the engine that turns the daily experience of a plant into standardised, reproducible knowledge. Its strength lies in the discipline of the closed loop, in verifying against hypotheses rather than against wishes, and in standardising each advance so that it is not lost. An organisation that masters PDCA learns faster than its competitors, and that pace of learning, sustained turn after turn, is the advantage that no rival can copy overnight. At Summum Quality we help manufacturing teams install this rhythm: starting from a real chronic problem, closing the first cycle with clean data and leaving the routine in place so that the team itself runs the second cycle.