Most improvement efforts stall in the same place. A team spots a problem, agrees on a fix, rolls it out and moves on to the next fire. Nobody measures whether the fix worked, so three months later the problem is back and the meeting starts again.
The PDCA cycle closes that gap. It forces two steps most teams skip: checking the result against a target set in advance, and turning what worked into the new standard. This guide explains the four steps, walks through a worked example in an office process, compares PDCA with PDSA and DMAIC and shows where the cycle sits in Lean and ISO 9001. Definitions come from ASQ, the Lean Enterprise Institute and The W. Edwards Deming Institute. The last section is our own angle: at Viralistic we run every automation project as a PDCA cycle, because an automated workflow is the most rigid standard a process can have, and you only lock in a standard you have tested.
What is the PDCA cycle?
The PDCA cycle is a repeating four-step method for carrying out change: Plan, Do, Check, Act. ASQ describes it as a four-step model that, like a circle, has no end, so you repeat it for continuous improvement. Each turn of the loop tests one change on a small scale, measures the effect and either standardises the change or feeds what you learned into the next plan.
The method goes by several names: the Deming cycle, the Deming wheel, the Shewhart cycle and, in Deming’s later version, PDSA (Plan-Do-Study-Act). In Lean it is the basic unit of problem solving. The Lean Enterprise Institute calls PDCA the foundation of continuous improvement, or kaizen.
| Step | The question it answers | What you have at the end |
|---|---|---|
| Plan | What is the problem, what causes it, what will we try? | A measurable target, a hypothesis and a small test plan |
| Do | Does the change work in practice, on a small scale? | A trial run with data collected as you go |
| Check | Did we reach the target, and what did we learn? | Results compared with the target, plus surprises noted |
| Act | Do we adopt, adjust or drop the change? | A new standard, or a sharper plan for the next cycle |
The four steps of the PDCA cycle
Plan takes the most thought, Do is deliberately small, Check compares facts with the target and Act decides what becomes standard. In Toyota’s own problem-solving training, more than half of the effort goes into Plan. If you understand the problem, the countermeasure tends to be obvious.
Plan: define the problem, the target and the test
In the Plan step you describe the current situation, find the root cause and decide which change to test. Write the problem down in numbers: how often it happens, where and what it costs. Set a target you can measure, such as “80% of quote requests answered within one working day”. Then state your hypothesis: if we change X, we expect Y. Decide in advance how you will measure Y, because without a target there is nothing to check later.
A process map is the fastest way to see where the time goes. A value stream map shows each step, the waiting between steps and where work is redone. Tools such as the 5 Whys or a fishbone diagram dig into causes.
Do: test the change on a small scale
In the Do step you carry out the change as a controlled trial: one team, one location, one product line or a fixed number of weeks. Keep the test small enough that a failure is cheap. Collect the data you defined in Plan while the trial runs, and note anything that does not go as expected. Do is a test, not a rollout.
Check: compare the results with the target
In the Check step you compare what happened with what you predicted. Did the change hit the target? Were there side effects? Is the improvement real or a lucky week? Base the judgement on the data from the trial, not on how the change felt. Deming preferred the word Study for this step, because it is about learning why the result came out as it did, not only whether it passed.
Act: standardise, adjust or start again
In the Act step you decide what to do with what you learned. If the change worked, make it the new standard: update the procedure, train the people involved and roll it out wider. If it worked partly, adjust and run another cycle. If it failed, drop it and plan a new hypothesis with what you now know. Either way, the next cycle starts from a higher baseline.
A PDCA cycle example, step by step
The clearest way to understand the PDCA cycle is to follow one through a real process. This example is illustrative, but the bottlenecks are the ones that come up again and again when we map office processes with clients: data entered twice and work that waits on one person.
An architecture studio with twelve people receives about 40 quote requests a month by email. Prospects complain that replies take too long, and some book a competitor in the meantime.
Plan. The team logs every request for two weeks. The median time from email to quote is three working days, and only 25% are answered within one day. Mapping the steps shows where the time goes: a request waits in a shared inbox, someone copies the details into a spreadsheet and then into a quote template, and the draft waits for a partner to approve it. Approval alone takes a day and a half on average, because the partners review quotes whenever they find time. Target: 80% of quote requests answered within one working day, within six weeks. Hypothesis: if partners review drafts in two fixed slots a day, at 11:00 and 16:00, waiting time drops sharply.
Do. One partner tries the fixed review slots for three weeks, for residential projects only. The team records the time each request arrives, the time the draft is ready and the time the quote goes out.
Check. In the trial, 64% of requests are answered within one working day, up from 25%. The median falls from three days to just over one. The target is not reached yet, and the data shows why: a third of the requests lack the floor area, so the studio has to email the prospect back and wait.
Act. Fixed review slots become the standard for all partners and are written into the studio’s quote procedure. The next cycle tackles the missing information: a short enquiry form on the website with floor area as a required field.
Second cycle. With the form in place, back-and-forth emails mostly disappear, and 83% of requests are answered within a day. Now the process is stable and measured. That is the moment to automate it, for example with an n8n workflow that turns the form into a draft quote. More on that below.
Notice what the studio did not do. It did not buy software in week one. It removed the waiting and the rework first, then automated what was left.
Where does PDCA come from? Shewhart, Deming and Japan
The idea behind PDCA comes from Walter A. Shewhart, a physicist and statistician at Bell Telephone Laboratories. In 1939 he described quality control as a cycle of specification, production and inspection, repeated to gain knowledge. His editor on that book was a young W. Edwards Deming.
Deming took the idea to Japan, where he taught statistical quality control to engineers and executives from 1950. He presented it as a wheel of design, production, sales and research. According to the Lean Enterprise Institute, the Union of Japanese Scientists and Engineers (JUSE) recast that wheel in 1951 into the Plan-Do-Check-Act cycle we know today, and in Japan it became known as the Deming cycle.
Deming himself never liked the word Check. Later in life he taught PDSA, Plan-Do-Study-Act, and in his 1993 book The New Economics called it the Shewhart cycle for learning and improvement. The W. Edwards Deming Institute explains that Check focuses on whether a change succeeded or failed, while Study asks you to predict the result, compare it with what happened and revise your theory.
PDCA vs PDSA vs DMAIC: which one should you use?
PDCA and PDSA are the same four-step loop with a different emphasis in step three, while DMAIC is a heavier five-phase method from Six Sigma for complex problems with a lot of variation. Use PDCA for frequent, small improvements, PDSA where learning and prediction matter most, and DMAIC when the cause is unclear and you need statistics to find it.
| Aspect | PDCA | PDSA | DMAIC |
|---|---|---|---|
| Steps | Plan, Do, Check, Act | Plan, Do, Study, Act | Define, Measure, Analyze, Improve, Control |
| Origin | Japan, early 1950s, from Deming’s teaching | Deming, published in 1993 | Six Sigma, which began at Motorola in the 1980s |
| Emphasis | Test fast, standardise what works | Predict, study the gap, learn | Data, root causes and reducing variation |
| Typical length | Days to weeks | Days to weeks | Often several months |
| Who runs it | Any team, on its own work | Any team; common in healthcare | Trained Green Belts and Black Belts |
In practice many organisations use both. A Lean Six Sigma Black Belt leads DMAIC projects on the big, data-heavy problems, while teams run short PDCA cycles on everything else. DMAIC’s Improve and Control phases are themselves a PDCA loop in miniature.
What about SDCA?
SDCA (Standardise, Do, Check, Act) is PDCA’s partner. PDCA raises performance to a new level; SDCA holds it there by checking that people follow the new standard. Many Lean teams alternate: an SDCA cycle to stabilise, then a PDCA cycle to improve.
What are the 8 steps of the PDCA cycle?
The eight steps are Toyota’s practical problem-solving method, which breaks PDCA into finer parts: clarify the problem, break it down, set a target, analyse the root cause, develop countermeasures, see them through, monitor results and process, and standardise and share. The first five belong to Plan, the last three map onto Do, Check and Act. Toyota writes them on a single A3 sheet, and as AllAboutLean notes from Toyota’s own training, more than half of that sheet is Plan.
How does PDCA fit into Lean and Kaizen?
PDCA is the method, Kaizen is the mindset and Lean is the system around both. Kaizen means continuous improvement in small steps by everyone, every day. Each of those small improvements runs as a PDCA cycle. In the Toyota Production System, the loop shows up everywhere: in A3 reports, in daily huddles and in the way a team updates its standard work after an improvement. Our guide to Lean management covers the five principles that PDCA puts into practice.
The link with waste is direct. Most Plan steps in an office end up pointing at one of the 8 wastes of Lean: waiting for approval, re-entering data, searching for files, correcting errors, reports nobody reads. PDCA is how you remove them one at a time and prove each removal worked.
PDCA in ISO 9001 and other management systems
PDCA is built into the structure of ISO management system standards. ISO 9001 for quality, ISO 14001 for environment and ISO/IEC 27001 for information security share a common clause structure, which ISO calls the Harmonized Structure, and its clauses follow the PDCA logic:
| PDCA step | Clauses in the shared structure |
|---|---|
| Plan | 4 Context of the organisation, 5 Leadership, 6 Planning |
| Do | 7 Support, 8 Operation |
| Check | 9 Performance evaluation |
| Act | 10 Improvement |
This mapping is the common reading among auditors; some guides place clause 7 under Plan. The 2015 edition of ISO 9001 described the PDCA cycle explicitly in its introduction, as part of the process approach, and said it applies to single processes and to the management system as a whole. ISO published the sixth edition, ISO 9001:2026, on 16 September 2026. Its topics still run from context, leadership and planning through support and operation to performance evaluation and improvement, and certified organisations move to the new edition within their certification cycle.
For a business, the practical point is this: a team that documents its improvements as PDCA cycles already produces the evidence an ISO auditor asks for under performance evaluation and improvement.
When should you use the PDCA cycle?
Use the PDCA cycle when you want to change a process and need to know whether the change works before you commit to it. ASQ lists the typical moments: starting an improvement project, designing a new or improved process, defining a repetitive work process, planning data collection to verify a root cause, implementing any change and working towards continuous improvement.
It fits best when:
- the process repeats often enough to measure, daily or weekly;
- you can test a change on a small scale without risking the whole operation;
- the team that does the work can run the cycle itself.
PDCA is the wrong first tool in two situations. In an acute incident, contain the damage first and run PDCA afterwards to stop it happening again. And when a problem has many interacting causes and a lot of variation, a data-heavy DMAIC project finds the cause faster than repeated small trials.
Common PDCA mistakes
The most common mistake is running PD, PD, PD: planning and doing, then jumping to the next problem without Check or Act. Without those two steps, nothing becomes standard and the old way creeps back.
The others follow from it:
- No measurable target. If Plan does not say what success looks like, Check has nothing to compare against. “It feels faster” is not a measurement.
- Starting too big. One problem and one change per cycle. Change five things at once and you cannot tell which one worked.
- Skipping the small test. Rolling out a change everywhere in the Do step means a failure is expensive, and you lose the comparison.
- Not standardising. A successful change that lives only in one person’s head is gone after the next holiday. Write it into the procedure.
- Turning PDCA into a form. When a PDCA template becomes a compliance exercise, people fill in four boxes and nobody learns anything. The template is a thinking aid, not the goal.
- Stopping after one round. The point of the cycle is the next turn. Every Check should end with the question: what do we try next?
PDCA for automation projects: pilot, measure, then scale
Automation is where PDCA earns its keep in 2026. An automated workflow is standard work that runs itself, thousands of times, without a person noticing when something is off. So the rule we work by is simple to say and often ignored: automate in Act, not in Do. Only a process that has passed a Check deserves to be encoded in software. Automate a broken process and you produce the waste faster.
That is also how Viralistic runs automation: value first, prove it in production, measure it, then scale. Here is how those four phases map onto the PDCA cycle, as described on our value-first automation page.
Plan: map the process before anything is built
We sit with the people who do the work and map every step: what starts it, which systems they use, what they decide and what goes wrong. The map shows where data is entered twice and where work waits on one person. Then we write a success plan: which process, which outcome, which metrics and who owns them. Rule-based steps are marked for plain automation; only steps that need judgement, such as reading free text or drafting a reply, get an AI model.
Do: pilot one to five workflows, in shadow mode first
We build one to five workflows in n8n, with Claude for the judgement steps and Notion or the client’s own tools for data. Each one goes live in stages: in shadow mode it runs next to the person and only produces drafts, then it moves to assisted mode where a person approves each action with one click, and only the proven cases become fully automatic. This method is the one Marrallisa taught in the DataCamp webinar on turning a business process into an agentic workflow, and it is written out step by step in our guide to building an agentic workflow.
Check: measure against the success plan
At each stage we measure time saved and error rate against the targets in the success plan, and we show the result in a simple value dashboard. That tells you when a workflow can move from shadow to assisted to automatic, and it gives you the numbers for the next decision. This is also where human oversight matters. Toyota’s principle of Jidoka, automation with a human touch, means a machine stops and flags a problem so a person can fix the process. A human in the loop for AI agents is the same idea: the workflow pauses on low confidence or risky actions and asks.
Act: standardise, govern and scale in waves
When a workflow proves itself, it becomes the standard: documented, owned by a named person, with error handling, logging and access rules. Our reliability and governance principles cover what that takes. Then the next wave starts, faster than the first, because credentials, templates and governance are already in place. Our n8n rollout approach describes that scaling in waves. If you want someone to run this cycle with your team, an automation specialist who maps before building is the person to look for.
Who teaches this
Marrallisa Kreijkes, founder of Viralistic in Amsterdam and an official n8n Ambassador, teaches this way of working to teams and audiences. She presented the DataCamp webinar “How to turn a business process into an agentic workflow”, trained Volkswagen Pon Financial Services on getting found in AI, spoke on the main stage of n8n’s conference In The Loop in Berlin in October 2026 and gave a lecture at Anthropic, the company behind Claude. She also ran an n8n workshop at Maastricht University with the student associations MAIN and TCM. Teams at DataCamp, Volkswagen Pon, n8n and Anthropic learned from her. When you work with Viralistic, you work with the person they brought in to teach. Read more on Marrallisa’s speaker page.
The PDCA cycle was born in the factory, and it is still the most practical way to make an office process faster and to know it is faster: plan with a target, test small, check against the numbers, then standardise. Run that loop before you automate and the automation lands on a process worth keeping. Start with the principles of Lean management, or talk to us about your first cycle.
Frequently asked questions
What are the 4 steps of the PDCA cycle?
The four steps are Plan, Do, Check and Act. In Plan you define the problem, the target and a change to test. In Do you test that change on a small scale. In Check you compare the results with the target. In Act you standardise what worked or adjust and run the cycle again with a better plan.
What is the difference between PDCA and PDSA?
The only difference is the third step. PDCA uses Check, PDSA uses Study. Deming preferred PDSA because Check suggests a pass or fail inspection, while Study asks you to predict the result, compare it with what happened and learn why. In daily practice most teams use the two names interchangeably for the same loop.
Is PDCA part of Kaizen?
Yes. Kaizen is the philosophy of continuous improvement in small steps, and PDCA is the method each of those steps follows. The Lean Enterprise Institute calls PDCA the foundation of continuous improvement, or kaizen. In German-speaking companies the same idea runs under the name KVP, with PDCA as its working method.
In which stage of PDCA is the solution identified and implemented?
The solution is identified in Plan and implemented in Do. Plan covers root-cause analysis and choosing the countermeasure to test. Do carries it out, first on a small scale. Only after Check confirms that it works does the solution get implemented fully, as the new standard, in the Act step.
How long should a PDCA cycle take?
As short as the measurement allows. A small change at one workstation can be tested in a few days, while a change to an office process usually needs two to six weeks of data. Make Check long enough to give reliable numbers and short enough that the team still remembers why it started.
What happens if the Check step shows the change did not work?
Then the cycle still did its job, because you learned something cheaply. You do not roll the change out. In Act you record why it failed, revise your hypothesis and start a new cycle with a better plan. A failed test on a small scale costs far less than a failed rollout.
Can you use the PDCA cycle outside manufacturing?
Yes. PDCA works in any repeating process: healthcare, education, customer service, finance, marketing and software. ASQ’s best-known example is the Pearl River school district in New York, which used PDCA from the boardroom to the classroom and won a Baldrige quality award in 2001. Office processes are often the easiest place to start.
Should you automate a process before or after a PDCA cycle?
After. Run at least one PDCA cycle to remove waiting, rework and duplicate data entry, and automate only the process that passed Check. The automation build is then a PDCA cycle of its own: plan the workflow, pilot it in shadow mode, measure the results, then standardise and scale.
Written by
Founder of Viralistic and official n8n Ambassador
Marrallisa leads strategy, design, search and automation for premium brands from Amsterdam. She has trained teams at Volkswagen and speaks at n8n and Anthropic events. Want to talk through your situation? Book a call with her directly.