QM BenchQuality engineering tools

The red bead experiment: is it the worker or the system?

Six willing workers produce white beads for a demanding customer. The box contains 20 % red beads, and each worker scoops 50 beads a day with the same paddle. The foreman ranks, praises, warns and finally fires people. Run the days and let the np chart tell you who is really responsible.

Day 0 / 4 · box 20 % red · paddle 50
Red beads per worker and day
WorkerDay 1Day 2Day 3Day 4Rank today
Anna·····
Ben·····
Carlos·····
Diana·····
Erik·····
Fatima·····
np chart (red beads per paddle)
Day 1Day 2Day 3Day 4
Mean red beads
0.0
Points outside limits
–
Rank correlation day 1 vs last
–
Expected by the box
10.0

Limits are calculated from the days shown: n·p̄ ± 3·√(n·p̄·(1 − p̄)). For the box alone they would be 10.0 ± 8.5.

Five guided lessons

  1. Four days of production. Run the classic experiment and read the np chart.
  2. The worker of the month. Check whether “the best worker” stays the best.
  3. Fire the three worst. See what happens when management “fixes” the people.
  4. Change the system. Improve the incoming material instead of the people.
  5. More inspection?. Understand what inspection can and cannot do.

How the original experiment works

Deming ran the experiment with volunteers from the audience: six willing workers, two inspectors, a chief inspector, a recorder and a foreman. The material is a box of about 4,000 beads, 20 % of them red. The procedure is strict: mix the beads, dip the paddle, tilt it, carry it to the inspectors. Workers may not change the method. Management announces a target of no more than a few red beads, puts up slogans and runs a merit programme.

After each day the foreman praises the worker with the fewest red beads and puts the worst on probation. After day 2 or 3 the worst workers are fired and the best ones work double shifts. Nothing improves, because nothing in the system has changed.

What the numbers say

ItemFormulaBox 20 %, paddle 50
Expected red beadsn·p10
Standard deviation√(n·p·(1 − p))2.83
Upper control limitn·p + 3σ18.5
Lower control limitn·p − 3σ1.5

A worker with 5 red beads and one with 15 are both inside the limits. The difference looks large on the shop floor, but it is exactly what the system produces on its own.

Seven lessons for managers

  1. Most variation belongs to the system. Deming estimated that 94 % of problems are caused by the system and only 6 % by special causes.
  2. Rankings of people inside the limits are random. The best worker today is not the best tomorrow.
  3. Targets and slogans do not change the box. They only add fear (point 10 of the 14 points: eliminate slogans and targets for the workforce).
  4. Firing people does not fix a system problem. The new team gets the same result.
  5. Inspection does not create quality. It measures the red beads; it does not remove them from the box (point 3).
  6. Improvement comes from changing inputs and methods: better supplier material, a better process, better equipment.
  7. A control chart separates the two. A point outside the limits is worth investigating at the worker’s station; points inside are management’s job.

Where you see red beads in a plant

  • Monthly scrap ranking of shifts or operators, with “worst shift” meetings.
  • Incoming material with a defect rate the operator cannot influence, blamed on assembly.
  • Audit scores compared between auditors or lines without a look at the variation.
  • Bonuses for “zero defects” months on a process with a stable 1 % scrap rate.

Frequently asked questions

What is the red bead experiment?

A teaching experiment W. Edwards Deming used in his four-day seminars. Willing workers dip a paddle with 50 holes into a box of white and red beads; red beads are defects. The foreman praises, warns and fires workers based on their red beads, although the result is set entirely by the box, the paddle and the method. The control chart shows a stable system.

What is the lesson of the red bead experiment?

Variation in results comes mostly from the system, not from the people. Ranking, rewarding or punishing workers for common-cause variation is unfair and changes nothing. Only management can improve the system: material, method, equipment, training and supplier quality.

What are the control limits in the red bead experiment?

With 20 % red beads and a paddle of 50 the np chart has a centre line of 10 and limits 10 ± 3·√(50·0.2·0.8) = 10 ± 8.5, so about 1.5 to 18.5 red beads. Any result from 2 to 18 is the system talking, not the worker.

Why is the average in a real experiment often not exactly 20 %?

In Deming’s physical experiments the long-run average differed from the share of red beads in the box, because the beads differ slightly in size and weight and the paddle is not a perfect random sampler. He used this to show that the mechanical sampling method itself is part of the system. This simulator draws perfectly at random.

How does the red bead experiment relate to performance reviews?

Deming listed annual merit rating among the “deadly diseases” of management. If most of the variation is common cause, a ranking of people is largely a lottery, and it destroys teamwork. The red bead experiment makes that visible in 20 minutes.

Can I use this simulation in a class or training?

Yes. It is free and needs no sign-up. Run the lessons together, let participants predict the next day, then show the np chart and the rank correlation. The quiz at the end works as a short test.

Test yourself and go further