Process Yield Statistics and Distributions

Estimated reading time: 1 minute

The costs of poor quality include all expenses incurred for not making or providing a perfect product the first time, including scrap, rework, re-purchasing raw materials, labor, and inventory¹. Companies operating at a three-sigma quality level can spend about 25% of their annual sales remediating poor quality costs². Other estimates put the costs of poor quality in the range of 25–40%³. Poor quality can destroy a company.

Unexpected product failures significantly increase sales, general, and administrative (SG&A) costs and lead to increases in inventories and fixed assets required to support operations. These indirect costs erode profitability more than those directly attributable to warranty claims processes. Product recalls negatively impact businesses financially and result in adverse publicity. Customers expect printed circuit boards to meet their specifications.

An internal Motorola study found that units reworked in production often encountered problems during early customer use, even though the defects identified were corrected during production. Rework can frequently stress a unit non-standardly and predispose a product to early failure⁴. Denson found a similar occurrence in semiconductor manufacturing, where the reliability of computer chips was statistically correlated with the die yield⁵. These studies suggest that circumstances in detecting and reworking defects in some units may produce undetected damage on other units. When delivered to customers, these weaknesses often contribute to early failures.

Defects are any items that exhibit a departure from specifications. A defect does not necessarily mean that the product cannot be used; it only indicates that the product result is not as intended. In essence, defects refer to quality characteristics. Generally, the count of defects is assumed to follow a Poisson distribution.

Defectives are units that are considered completely unacceptable for use. Each unit is deemed defective or not—there are only two choices. In essence, a defective unit refers to the overall product. Generally, the count of defective units is assumed to follow a binomial distribution.

To read this entire article, which appeared in the November 2024 issue of PCB007 Magazine, click here.