Acceptance Sampling for Quality Protection

The BIS.Net Team BIS.Net Team

There has been considerable controversy over the use and benefits of Acceptance Sampling. The objective of this small article is to provide a balanced perspective for practitioners to decide for themselves whether to apply or not apply acceptance sampling as an additional form of quality protection. This will be achieved by reviewing the history of acceptance sampling and technology and the driving forces behind the controversy.

Additionally, the technology used to derive Acceptance Sampling Plans has to an extent led to lesser use of Acceptance Sampling. This article will discuss some of the reasons.

First, it impossible to precisely give a historic account because history has an infinite horizon of events. All that can be done is provide an account of general perceptions, much driven by historian’s perceptions. In this instance the historians are mainly notable Quality Gurus, such as Deming, Crosby, Feigenbaum and countless experts.

The general viewpoint is that in the fifties the function of Quality was one of inspection. A specialist team of inspectors would inspect product at the end of the production line and reject or accept the product. Little attention was placed in the process to prevent defect formation. (In fact process Control had already been introduced by Shewhart twenty years before.)

In the seventies emphasis was placed on process control. Inspection continued but processes were also controlled using statistical methods.

The eighties saw many philosophical changes in the role of the quality function. It was no longer considered appropriate for a specialized team of quality inspectors to be used. The responsibility of quality fell into the production department. It was reasoned that production manufactured the product and hence should be responsible for the product. There should not be a need for a ‘police’ department.

It was reasoned that there should be no need for incoming inspection. A better alternative is to work with the supplier and ensure that they had a sound quality system in place and then audit the system.

Much of the pressure to adopt the new way of thinking came from government bodies who would only deal with certified suppliers, likewise the powerful automotive companies, and consultants and academia. The result was less reliance on acceptance sampling and more on getting it right in the first place.

The philosophy behind the new way of thinking is hard to fault. It makes sense to get it right first. If you get it right first defective product will not be produced. It makes sense to rely on sound systems. The systems will ensure you get it right first. There should be no need to spend money on inspections and sampling after manufacture of product.

This philosophy is no matter how well meaning, one sided in that it places emphasis on the process until completion of the products, but not the final product. Overall it is safe to say that the most balanced approach was in the seventies where emphasis was placed on process control, but inspection continued. Missing however, was the Quality System Approach introduced later which is one of the most important components for ensuring the highest level of outgoing quality.

The eighties were more about a philosophical approach of getting it right first using quality systems. It is however easy to fall into the trap of being blinded by an idea that sounds infallible. However, the new way of thinking is not infallible. Systems fail, systems are not always strictly adhered to. SPC is not always applied diligently, things can go wrong between SPC sampling, especially with a low sampling frequency. Automatic systems are not always reliable.

Acceptance sampling is also not perfect. There are problems of taking random samples from a lot because it is not easy to reach every item stacked on pallets. There are sampling risks. But it does provide an extra level of protection and is probably cheaper than the cost of maintaining systems in place. The additional cost may be worth the extra level of protection it provides, but of course this depends on each situation.

For incoming goods, the cost of sampling in many cases can be far less than the time required to work with suppliers to set up and audit their systems. Suppliers are trying to make a profit. Suppliers are human and can be inclined to hide defects in their system from auditors. The writer has witnessed this several times, where records were hidden from auditors, or quality certificates were falsified.

For those who see benefits in Acceptance Sampling, computing sampling plans has in the distant past been difficult to the sampler. This has according to the writer’s experience resulted in a reluctance towards using formal Acceptance Sampling. Instead unreliable ad-hoc sampling plans were used, e.g. take a sample of 5 and reject if 1 or more defective samples are found.

There exist a vast number of tables such as the Dodge and Romig tables.

Acceptance Sampling Figure 1 - Dodge and Romig plans

These however were only designed for Rectification Attributes Plans. Additionally, there are several comprehensive tables produced by various standards organization and the US Military. All these tables are for some difficult to understand and thus risk being misapplied. The US Military and similar sampling plans were indexed by AQL only, i.e. the acceptable quality level and not rejectable quality level. The user would have to peruse OC curves to ensure the sampling plan provides adequate protection.

Acceptance Sampling in the modern era

Today it is easier to obtain sampling plans using computing power, but even with the use of computing power the best sampling plans are not always available, depending on how the software computes sampling plans. Some still use the methods used by Dodge and Romig, relying on approximations, but these approximations are not as accurate as was assumed at the time.

For example, using Machine Power to obtain ‘exact’ sampling plans for the case of 5% LTPD with rectification with a Consumer Risk of 10% we present the following comparison.

Acceptance Sampling Figure 2

Where n is the sample size, c the acceptance value and EUI the expected number of units inspected.


Whether a business decides to use acceptance sampling as a final layer of protection is its decision. What is important is that the decision is based on a balanced viewpoint considering the cost of applying acceptance sampling and the cost of not applying. Equally important is understanding that no system and philosophy is fail safe, no matter how much it makes sense.

200 Analytics Tools For Quality

Download 200 statistical analytics tools for quality improvement - Mobile Version

Improve product quality through smart data insights using new-age, machine-powered driven analytics, available to download through a suite of Apps!

  • Apps for SPC, MSA, Process Performance, Inferences, Visualization, and much more
  • No licencing or subscriptions! Pay ONLY per analysis billed monthly - don't use, don't pay!
  • Always up-to-date with the latest features