# What is the best practice when you have zero defects?

I recently had a conversation
with a customer who was collecting defect data each shift and charting it as
a percent. The “problem” was that for this chart, she had
ten consecutive days with zero defects. The thirty samples (10 days x 3
shifts/day), shows a flat line and no control limits. Since
there is no variability in the data, the standard deviation is 0, therefore control
limits can’t be calculated. She further explained that they are “required” to
chart this data. What would you recommend she do in this case?

I can think of a few options such as:

1. Ignore the ‘requirement’ to chart this since charting this does not add value.

2. Change the time period how defects are counted. For example instead of each shift, count them each day.

3. Show the control chart with no control limits and all of the data at 0.

4. Do something else.

Please share your suggestions.

## Comments

Understanding SPC, "When events become rare, the counts will be inherently insensitive to changes in the process". So rather than attempting to chart something based on a count of your rare events, you may find more success by charting the amount of time, or number of conforming products that pass between defects. These can be done with a t-chart, or g-chart respectively.1) Does the measurement system have the precision and accuracy at least 4x greater than the tolerance limits? If the variation is less than the accuracy of the measurement instrument it's possible to have the same reading every time. (Example tolerance/range of process is ± .0002" and a caliper accurate to ± .001" is used).

2) Are the values recorded in the full decimal place? If the variation is low, say ± .0004" and the spreadsheet is set up for 2 decimal places it will always be the same value.

3) Did they perform a GR&R and is the NDC value greater than 5?

4) Have people performing the input been trained on what's expected?