QM BenchQuality engineering tools

Find the defects: how good is 100 % visual inspection?

You are the final inspector. Each round shows a batch of plates with two holes; some have a scratch, a dent, a missing hole or a burr. Click every defective part before the time runs out. Five rounds change the number of parts, the time and how rare the defects are.

Round 1: easy start

Few parts and plenty of time.

24 parts · 25 seconds · click every part with a defect (scratch, dent, missing hole, burr on the edge).

The parts are hidden until you start. Good parts have two holes, smooth edges and a clean surface.

What the game shows

  • Nobody finds everything. Detection rates of 70 to 90 % are typical, and they vary between people and rounds.
  • Rare defects are missed more often (round 3), because the eye stops expecting them.
  • Time pressure costs detection and adds false alarms (round 4), like a faster line or the end of a long shift.
  • A second check helps but never reaches 100 % (round 5).

What to do in a real plant

Check how well your inspectors agree with an attribute agreement analysis, plan inspections with the inspection plan builder and choose sample sizes with the AQL calculator. Above all, prevent the defect: see poka-yoke and the quality inspection hub.

Frequently asked questions

How effective is 100 % visual inspection?

Much less than people expect. Studies of human inspection commonly report that inspectors find around 80 % of defects, with large differences between people, defect types and conditions. Rare defects, short cycle times and long shifts lower the rate further.

Why are rare defects harder to find?

When almost every part is good, the eye and the brain stop expecting a defect. This is the prevalence effect: the rarer the defect, the more often it is missed, even by experienced inspectors.

Does a second inspection solve the problem?

It helps, but less than the arithmetic suggests. Two independent inspectors who each find 80 % would together find 96 %. In practice the second inspector is not independent: both tend to miss the same subtle defects, and each relies on the other.

What works better than more inspection?

Preventing the defect: error-proofing (poka-yoke), capable processes and control of the process inputs. Where visual inspection stays, use limit samples, good lighting, short inspection periods, rotation and an attribute agreement study to check the inspectors.

Can I use the game in a training?

Yes. It is free and needs no sign-up. Let participants play the five rounds, collect their detection rates on a flip chart and discuss the spread between people and rounds.