May 4, 2011
Using Acceptance Sampling to Improve Quality
How do organizations confirm that supplier products comply with critical quality standards? For the most part, companies rely on inspectors to check incoming materials. Their results are compared with the company’s own quality standards and supplier-generated documents such as Certificates of Analysis (COA’s). Based on results, inspectors either accept or reject a shipment.
Organizations typically use Acceptance Sampling procedures as defined by the old MIL STD 105/414 or the more modern ANSI/ASQ Z1.4 or Z1.9. Regardless, Acceptance Sampling helps minimize inspection costs, manage risk, and prevent off-quality product from entering the production process.
But, most quality professionals view Acceptance Sampling as just another inspection tool. Most of us can recite the 3rd of Dr. W. Edwards Deming’s 14 management principles which exhorts organizations to cease reliance on inspection. Personally, I believe it’s best to expend resources on inspection of critical received goods rather than not. By not inspecting, companies are basically crossing their corporate fingers and hoping that products meet requirements.
Acceptance Sampling occupies the middle ground between no inspection and 100% inspection. The result is that these techniques have been derided as just another set of inspection tools. Plus, most quality professionals deem Acceptance Sampling as unworthy of being labeled as a quality improvement tool because the end result of all those statistical gyrations is a meek, stand-alone “Accept” or “Reject” conclusion. Long ago I agreed with these assertions, but no more. Instead, I believe Acceptance Sampling can be used as a highly effective means of improving quality.
Here’s how: Imagine inspectors use Acceptance Sampling to check incoming product. But assume inspectors save the data which has been used for making the Accept/Reject conclusion. For example, when performing attributes checks using ANSI/ASQ Z1.4, the actual defect codes and reasons for failure might be noted along with the supplier name, product code, lot number and other important traceability fields. Likewise, when using ANSI/ASQ Z1.9 for variables data, the actual measurements (and traceability elements associated to the shipment) are saved to a database. By doing so, not only would an Accept/Reject decision be made, but the Acceptance Sampling plan itself would allow data which lead up to the conclusion to be saved to a database.
Therefore, historical data is available by supplier, product and other traceability elements. The Acceptance Sampling plans themselves become the source for these quality data. If so, control charts, histograms, Pareto charts and other statistical analyses could be used to analyze receiving inspection data. Defect levels between suppliers could be compared. Significant time-based changes in PPM defect rates could be identified. The control (or lack of control) of a supplier’s processes could be confirmed. This data could be used to collaboratively work with the supplier to help the vendor improve the quality of their supplied products and their manufacturing processes.
In essence, saving the measurement data along with the Accept/Reject conclusion would allow Acceptance Sampling procedures to be a tremendous compliment to typical quality improvement efforts. Previously unknown vendor-specific quality levels could be accurately quantified and that information could be used for significantly improving vendor quality and reducing quality costs across the supply chain. And it can all be done simply with little or no cost to those who already use ANSI/ASQ procedures. All it takes is a change in mindset. We must think differently about Acceptance Sampling.