Using PDCA for Small Improvements: Short Loops and a Check Step You Don't Skip

PDCA is a loop for testing an idea and learning from it

PDCA stands for Plan, Do, Check, Act. You repeat the four steps to improve how you work, a little at a time. Its core is that a plan is not finished when you carry it out; you look at what happened and use it in the next round.

The roots are usually traced to statistician Walter Shewhart, who described a cycle of specification, production, and inspection in his quality-control work. W. Edwards Deming spread the idea, and in postwar Japan it became known as the Deming cycle. Later, in The New Economics (1993), Deming preferred the label PDSA, replacing Check with Study. The point was that checking is not enough; you have to think about what you learned.

Running it on small improvements

Applied to a big plan, one loop can take months, and learning slows to match. For a one-person business or small company, a loop of one to four weeks is more realistic.

• Plan: Pick one thing to fix. Write a hypothesis ("if I do X, Y should happen") and the number you will use to check it.
• Do: Try it on a small scale. Do not change everything at once.
• Check: Compare that number with the value from before.
• Act: If it worked, make it the new standard. If not, drop it or adjust the hypothesis and go around again.

Say replies to quote requests are slow. Your hypothesis might be that a reusable reply template will shorten the time to respond. The number is the average time from inquiry to reply. Because you wrote both down first, you can check later.

How it turns into ritual

PDCA is often reported to degrade into a formality in these ways:

• The Plan is a slogan. There is no hypothesis, so nothing is being tested.
• The Check is skipped. You did something but never compared it with before.
• The measure is chosen afterward, so any outcome can be called a success.
• The Act is vague. "Keep trying" changes nothing about how the work is done.
• The loop is too long. By the time you check, the conditions have shifted.

A missing Check is the most common. Doing is visible; checking is not, so it slips unless it is scheduled. Put the review date on your calendar on the same day you start the trial.

Where it fits poorly

PDCA suits work you repeat. If every job is different, or if you are exploring without a known target, there is little to state as a hypothesis in advance. In that case, try something small first and write down what you learn.

Small samples also cause trouble. With few data points, you cannot tell chance from improvement. When counts are low, avoid judging by one round; run two or three and look at the trend.

An example of writing the hypothesis

A hypothesis is easier to write when it has three parts: a condition, an action, and an expected change. For example: "If I reply to quote requests on the day they arrive, using a template, the number of days to close should shrink." Make the change measurable. "It gets better" cannot be checked.

Choose one number. If you watch several, at least one will probably improve, and you will be tempted to call it a success. Fixing a single measure in advance prevents choosing a flattering number afterward.

Deciding whether to keep or drop

The hard case in the Act step is a small effect. When the difference is small, run one more loop and look at the trend. If two loops show nothing, you can drop the change. Leaving one line about why you dropped it keeps you from repeating the same attempt. As the notes accumulate, they show you the pattern of how your own work behaves.

Keep the record to one spreadsheet row per loop: date, the issue, the hypothesis, the measure, the result, and the next decision. Six months later you can look up what you tried and what did not work. When you start delegating, the same record becomes a first draft of a procedure.

What to do next

1. Choose one thing to fix, such as slow replies or missed invoices.
2. Write a one-line hypothesis and the number that will test it.
3. Record the current value before you change anything.
4. Set a trial of one to two weeks and put the review date on your calendar.
5. On that date, compare before and after, and write down whether you will keep, drop, or revise the change.