Inspection of attributes is a chronic problem for many organizations. The more common attributes inspected are color, scratches, and presence or absence of a required item. They are counted in whole numbers, there is no such thing as 1.5 defects, it is either a defect or it is not. They can be random in nature, sometimes the defect is present and sometimes not. They can also change locations and characteristics, a scratch for example. This randomization renders vision inspection technology impractical, leaving the inspection to workers within the value stream.
Attribute inspection can cause operator fatigue and therefore mistakes. This is because a large number of samples are required in order to make confident decisions or 100% inspection is sometimes compulsory. Another contributor to the fatigue is in how the limits of the attribute are set or described. With variable data there is a well-defined upper and lower specification making the decision simpler but with attribute how do you define something like scratches or color (if a measuring device is not practical)? This lack of definition causes frustration which again, contributes and leads to fatigue.
In hypothesis testing the null hypothesis must be well-defined in order to prevent making a Type I or Type II error. A Type I error can be basically defined as rejecting the null hypothesis when it is actually true. A Type II error is defined as rejecting the alternative hypothesis when it is actually true. To have confidence in the decision a scientific study is conducted and the result is called statistically significant, it did not occur by chance. In the process of quality inspection of attributes we do not have this luxury, data is not gathered by the inspector, and decisions are made within seconds. In my personal experience of conducting gage Reproducibility and Repeatability studies the outcome of this inspection process is commonly 65% Type I errors, a non-defective unit is rejected. As in hypothesis testing we can raise the question of which error is worst infographiczon, a Type I where a good unit is scrapped or a Type II where a defective unit is accepted and likely passed to the customer. It comes down to cost. If rejecting a good unit is low cost then an organization would probably prefer Type I errors. Type II errors in this case would likely pass on the poor quality to a customer and we all know how expensive a lost customer can be.
Of course, as Lean Sigma Practitioners we prefer to take the preventative approach and eradicate the root causes of these defective attributes. But as I said earlier defective attributes continue as a prolonged problem in many organizations. So in the interim we need to look at different inspection practices and processes to ensure protection for our customers. As in hypothesis testing there must be a significance level, something the inspector can confidently base their fast paced decision on, is it good or bad. This “significance level” can only come from the customer, the party we are trying to protect. We do not have data to determine our confidence but data does not always determine if our decision is correct or not. For example, if I am an inspector looking at a borderline scratch and I know this customer complained about scratches from their last order, my decision is to scrap this unit. This is a fast paced “customer driven” decision by an inspector.
So in the interim here are some proven inspection practices and processes you can put into place to safeguard the attributes you are sending your customers.
Define – Back to my reference of hypothesis testing, the null hypothesis must be well-defined, it is the reference. The inspection process must also have well-defined references in order to make the correct decision. For the unit being inspected, break it down into a grid with critical, major, and minor zones. Simplify the decision by stating attributes cannot vary within a critical zone. Infographics is very powerful for this application because it combines visual graphics and knowledge to define a defective attribute.
Source Inspection – Start by determining the process steps these defective attributes are coming from. This is where inspection must take place. I have made reference to inspectors but these are not end of line inspectors, they are people running the value add processes. End of line inspection does not even have a remote chance of revealing these long time root causes. Understanding the mechanics of the process provides inspection with knowledge about the possibilities of where these defective attributes are coming from. Have these possibilities visually managed for further discussions during Gemba walks or Kaizen events.
Visual Management – Do not put defective attribute references in a book or poorer yet, stored in a computer file folder. Medium like these do not have enough exposure and without exposure people will not learn so it becomes automatic at decision time. Visual management of the critical few and the more chronic problems will get people’s attention. The visual management must be larger than what is normal for visual management formats.
The daily management Gemba walk – Use this practice to have managers review the visual management with the inspectors. Be prepared with questions to ask the inspectors about what they are encountering and the decisions they have made. Make sure they understand the visually managed references. Taking samples from their reject bin can also support the required learning of the references. Finally, the daily Gemba walk can also be used to reinforce any customer issues related to defective attributes.
Internal supplier and customer process – Design your value stream with defined internal suppliers and internal customers. Each process step team, as an internal customer to the previous process step defines their acceptance of quality. They are responsible to collect data on any defective attributes they are receiving. At the end of the week the internal customer has a data driven discussion with their internal supplier. The internal supplier then has the responsibility to improve any defective attributes. These actions are presented at the next weekly meeting. This will interconnect all of the value add steps to improve quality and safeguard the external customers.
Sales Team Involvement in The Gemba – If there is a customer concern (should not wait until a formal complaint) about quality a person from the sales team must visit the value add process step which most likely caused it. There is a quick review of the visually managed reference and the visually managed possibilities of causes. There might even be a Kaizen event started to understand if they can reveal the root causes and provide countermeasures. Even if there is no Kaizen event these workers are now enabled with customer driven knowledge.
Self-directed team organization – With this type of organization the self-directed teams are empowered with customer knowledge, they are cross-trained to prevent fatigue creating tasks, and they are more developed in problem solving skills to expose these elusive root causes. They are the closest people you have to these root causes.
There are other changes which can be made to process and practice to improve attribute inspection. Please provide your comments on what you have experienced.