PDCA Cycle
Plan · Do · Check · Act
What is the PDCA cycle?
- Plan-Do-Check-Act, repeated
- Try small, learn, and improve
- The backbone of continuous improvement (kaizen)
What PDCA really changes is the unit of execution. Instead of treating 'this project' as one big block, PDCA makes a two-week or one-month cycle the unit of work. The key difference is that the review date is fixed on the calendar in advance. If you don't decide what you'll review and when you'll make the call, the review turns into after-the-fact reflection, and the plan hardens without ever being revised. Setting the decision date first is half of PDCA.
Of the four steps, C is almost always the weakest in practice. But the cause of a weak C usually lies in P. If P doesn't pin down numbers and a baseline — not 'reduce misdeliveries' but 'cut 38 a month to 15 or fewer' — the review meeting ends with 'it seems better than before.' A also needs to be split in two. What has proven effective is locked in as standards and checklists so it doesn't slide back, and only what hasn't been confirmed is passed to the next P. The boundaries are clear too. PDCA is the right tool for improvements where goals and measures are already in place. For a new business where customer response is itself unknown, a hypothesis-testing approach fits better, and forcing PDCA onto it only makes the planning documents thicker.
The most common failure is letting PDCA harden into a reporting form. Once a four-box form is filled in at month-end and sent up the chain, 'Progressing as planned' gets copied into the Check box and 'Will continue to monitor' into the Act box every month. If the owner gets scolded for honestly writing down a failed Do, the numbers get polished from the next cycle on, and from then on the cycle keeps turning while nothing improves. Cycles that are too long are a problem too. Run it once every six months and you get only two chances a year to learn, and if the owner changes in between, the previous cycle’s records become someone else’s document. It is far better to keep each loop short and keep a record so the next person can pick it up.
John KimAuthor of B2B Marketing Design and CEO of Nanum Consulting
Naha JinStartup leadership specialist with a social welfare backgroundWalter Shewhart's idea was developed and popularized by W. Edwards Deming, becoming widely known as 'PDCA (the Deming Cycle).'
Let's look at a hypothetical case. Company A, a food processor with 240 employees, ships an average of 1,200 orders a day from its logistics center in the Seoul metropolitan area. In the second half of last year, misdeliveries and quantity errors rose to 38 a month, and client complaints and return freight alone were costing KRW 9 million (≈ US$6,730) a month. The head of logistics said, 'Let's cut that in half within 12 weeks,' and decided to run PDCA in two-week cycles.
- Plan (P) — The logistics manager broke down the past six months' 152 errors by type and found that 63% were picking mistakes between similar items. The goal was pinned to a number: 'cut 38 a month to 15 or fewer within 8 weeks.'
- Do (D) — For the first two weeks, they focused on the 20 items with the most errors, color-coded their shelf labels, and piloted a two-person cross-check with four night-shift workers.
- Check (C) — Two weeks later, errors had dropped from 38 to 27, but the check also revealed that the cross-check had pushed closing time back by 40 minutes. The night-shift foreman said, 'People can't sustain this pace.'
- Act (A) — They limited the cross-check to the 20 color-labeled items and changed the system so the ship button stays locked until a PDA scan is done, then moved on to cycle 3.
- After four cycles over 12 weeks, errors fell from 38 to 9 a month, and return freight dropped from KRW 9 million (≈ US$6,730) to KRW 2.1 million (≈ US$1,570) a month. Closing delays returned to normal levels, and the confirmed procedure was codified in a three-page work standard.