Skip to content
Quality Tools | Part 2 of 8

Check Sheet and Stratification: Count First, Then Sort

How to define defects, count them on a check sheet and sort the data by shift, position and weekday until a pattern shows.

A check sheet is a prepared form on which each occurrence gets one mark. It sounds trivial, and it is the tool most often done badly. Counts are reconstructed from memory at the end of the week, defect names mean different things to different inspectors, and nobody records how many units were inspected. Then the data cannot answer any question.

Define before you count

On powder coating line 2, five decisions were written down before the first mark:

  • The unit. One coated part after curing.
  • The defect types. Seven, each with a photo and a limit sample at the inspection station: thin coating, inclusions, orange peel, scratches and handling damage, craters and pinholes, color deviation, other.
  • The counting rule. A part with several defects is counted once, under its main defect, the one that decides the rework.
  • Who and where. The inspector, at the inspection station, one mark per part.
  • The factors. Date, shift and hanging row of the part.

The last point matters most for what follows. A factor that is not on the check sheet cannot be sorted later.

Before the count started, two inspectors classified the same 20 sample parts independently. They agreed on 15. After the limit samples for orange peel and inclusions were renewed, they agreed on 19. Below 18 of 20, sharpen the definitions before you collect anything.

Four weeks of counting

Twenty days gave 24,000 inspected parts and 1,920 marks.

Defect type Parts
Thin coating 720
Inclusions 480
Orange peel 288
Scratches and handling damage 192
Craters and pinholes 120
Color deviation 72
Other 48

Stratification: the cheapest analysis there is

Stratification means sorting the same data by a factor. A difference between the groups points toward the cause. No difference rules a factor out, which is just as valuable.

The team sorted the 720 parts with thin coating by shift first, because the shift was the favorite explanation in the plant. The early shift had 352, the late shift 368. With 12,000 parts each, that is 2.9 and 3.1 percent. No difference. One long discussion was over.

Then they sorted by hanging row:

Hanging row Parts inspected Thin coating Rate
Top 8,000 144 1.8 %
Middle 8,000 168 2.1 %
Bottom 8,000 408 5.1 %

A part on the bottom row is almost three times as likely to be thin as a part on the top row. The bottom row carries 56.7 percent of this defect.

The 480 parts with inclusions showed nothing by shift or by row. By weekday they did: Monday had 168, the other four days between 74 and 84. With 4,800 parts per weekday, Monday runs at 3.5 percent and the rest of the week at about 1.6 percent. The line stands over the weekend, and dust settles in the oven entrance and on the conveyor.

Compare rates, not counts

A group that makes more parts has more defects. Always divide by the units inspected in the group before comparing. A simple index helps: the rate of the group divided by the rate of all groups. Thin coating runs at 3.0 percent overall, so the bottom row has an index of 1.7. Inclusions run at 2.0 percent overall, so Monday has an index of 1.75. Above 1.5, look closer.

Write down what showed no difference

On line 2 the powder lot was a second favorite suspect. Sorted by the six lots of the period, thin coating ran between 2.8 and 3.2 percent. No difference. Shift and powder lot went onto the wall as rejected factors, with the numbers. Without that note, the same suspicion returns at the next meeting.

Newsletter

One Lean idea.
Every two weeks.