Learn more effective ways to inspect than AQL... You can’t inspect quality into a product. [Podcast.ep.040]

 

Simple. Free. Advice. No Obligations. 

Podcast Transcript 

David Hoffmann 0:22
Hi, everybody, and welcome to this week's podcast. I've got Joyce here with me. Say hi, Joyce.

Joyce Ng 0:29
Hi. How's it going?

David Hoffmann 0:32
Oh, good. Oh, good. Today, we've got a great guest. We can't see what fortunately have some video problems. But we have Dr. John Higgins in the house. How are you Dr. John?

John Higgins 0:43
On very well. Thanks, David. And all going well, the sound of the world. Great. Where

David Hoffmann 0:49
are you, John?

John Higgins 0:50
I'm here in South Africa, Pretoria. Great.

David Hoffmann 0:54
And how's the Coronavirus analogy. New strain is how things going there.

John Higgins 1:00
Now, the latest strain is not concerning, actually, because the effects are very mild. As compared to the previous previous viruses. This one seems to be mild. people recover really quickly.

David Hoffmann 1:18
Yeah, I've heard that. That as well. Yeah, I've had a spread very rapidly, but it's not very serious.

John Higgins 1:25
It's not, it's not affecting the hospitalities and so forth. So yeah, it's it seems to last a week, and people are recovering. So it's it's quite, quite loud in that way.

David Hoffmann 1:43
Great. Great. Well, John, so just to give a quick introduction to our audience, and everyone listening about who you are, for those of you listening, Dr. John is a doctor of electrical engineering. And he's been a mentor and advisor to me for the last 2025 years on a number of projects. And if anybody in this world, who is the most experienced quality, quality control person I know, it's definitely Dr. John Higgins and our black belt expert in six sigma and lean systems. John, maybe tell people a little bit about your background and what you do.

John Higgins 2:23
Right. So I started off as a electronics engineer, covering radio, microwave TV cable transmission, and then studied further into electrical electronical quality engineering, and then posted into management to PhD level. And I was an executive for a major manufacturing company for a number of years covering products such as TV, Hi Fi, microwaves, stoves, fridges, freezers, and so forth, and audio products and gained a lot of practical experience in in Quality Engineering and Management. And as you know, other blood in your company and many other companies around the world in Saudi Arabia, Saudi Arabia, and all around South Africa. Yeah,

David Hoffmann 3:15
no, absolutely. You haven't. We Glad to have you on the podcast because I speak to you a lot about quality control and the principles in the background, as you know, and I've been thinking for a long time, that would be great to actually let people hear firsthand from somebody like you some of the conversations we have around quality and, and methodologies around managing quality. So John, a couple of just a quick question, I know that you involve now in doing six sigma training. And I've made some people may have heard of Six Sigma some may not in a nutshell what exactly is Six Sigma just so we can give it to kind of encapsulate that in in under a minute.

John Higgins 3:58
Okay, so there are two dimensions, one is lean and other one is six sigma, lean looks at improving the workplace from a Foggiest point of view. And various other techniques, we Six Sigma is most statistical, so that if you want to solve a problem, and it requires statistical analysis, six sigma would be the way to go. Lean is much more simpler, more, more easy to implement, but they work together. And that's the important point. I've been giving lectures on both Lean and Six Sigma radically black belt level all around the world. And the methodology is interesting. There are five dimensions to the methodology. So people learn how to solve problems by using the methodology of demek. And students are required to do not only theoretical tests, but they have to be practical projects to show that they understand and can apply practically

David Hoffmann 4:59
So John ffor the sake of our audience, I want to kind of explain him a little bit. I want to bring it down to it to a practical level. So go with me here. What are the five S's? Can you what can you run through them quickly?

John Higgins 5:15
Sort, set in order. shine, standardize, sustain. And forgot the 4, one. But anyway, those are the issues.

David Hoffmann 5:27
Okay, so so so the way it kind of practically looks in my understanding, and correct me where I'm wrong, is if you went into a manufacturing environment, for example, and this doesn't only apply to manufacturing and applies to even an office workspace or anywhere, actually, any business organisation that runs, you would look at the five S's. And it's kind of like a management process of how well things are organised, managed, controlled and executed. Is that kind of a failed way to look at it in a in a simple in simple terms?

John Higgins 6:01
Yes. So that's exactly correct. And if you get that correct, you're able to sustain a flow in your processes, and then address the defects and so forth, so that you'll get good customer service as quickly as possible.

David Hoffmann 6:19
Right? So so it's, it's that basically, if your business is run on, well done processes, you're going to be able to identify problems quickly and recover from the fact whether it's a manufacturing process, or even a customer service process. For example.

John Higgins 6:37
I always say to my customers, everything's a factory, whether it's a bank, or a factory, or whatever, everything's a bank, because it is a process a process. Yes.

David Hoffmann 6:51
Yeah, I agree. And I must say, when you taught me that, kind of a penny dropped for me is that business processes are like a factory. Customer service is like a factory, if you get that process, right. It's the same principles that apply everywhere. And I think that's what fascinated me for years around five essence, Six Sigma, and these Lean process management systems that they just apply in everything and anything you do, even in software development.

John Higgins 7:20
I'm currently helping a customer with his warehouse. And city Amalia warehouses effectively looked at me and said, I don't understand. I said, Well, it's very simple. Your marketing division presents the warehouse with a picklist. To deliver that big list is actually the production schedule. Yes, that drives the deliveries. And so that sort of dropped the pin in his mind. So now they're very prestigious thinking. And that's what you got to get right with, with, with your customers, if you're trying to help them improve the methodologies in whatever environment might find.

David Hoffmann 8:01
Yeah, and I must say that, the other thing you taught me, is always listened to the voice of the customer, the VRC. And what I understand is everything you do in your business, or in manufacturing, it's all driven by the voice of the customer. So if you kind of take that approach of what is the customer actually want, and then figure out how to develop a design a process or a product to deliver that, not the other way around, not design a product or business process and force customers to go, Well, this is the way we work.

John Higgins 8:35
Voice of the Customer is a very broad concept. So it's important to break it down into what we call critical customer requirements. So you take the high level, the voice of the customer, and you break it down into those critical customer requirements. And those are the things that the factory is was attempting very serious. In the inspection, when you

David Hoffmann 9:02
say factories, John, just interrupt you in, I mean, from our constituents, when you say the factories must actually it's deeper than that, right? It's whatever the voice of the customer is, the factory or the business or the organisation, whatever it is, should adapt and develop process around delivering that drive.

John Higgins 9:22
Absolutely today. And it's very important, those processes, address the critical customer requirements, they should measure them and do some statistical tests on those measurements to establish confidence that will express itself in a reliable product. And that

David Hoffmann 9:44
yeah, now it's fascinating when you think about it like that it actually makes decision making very simple and clear, doesn't it?

John Higgins 9:51
That's very important. People tend to complicate things, and I always break it down and make it as simple as possible so that people people grasp it and then run with it. Otherwise, they get lost and they don't achieve what they should.

David Hoffmann 10:07
Yeah, I mean, I find I just see it repeatedly in Mark company in other people's companies. And it's not that it's wrong, it's just the cycle of development and improvement processes, is, you get so caught up in doing things, because you're used to doing in that way. Whereas if you take a step back and go, What is that customer really wanted the end of the day, or what the product actually do at the end of the day. And when you start thinking like that, then you decide what you need to do. And you can actually strip a lot of steps out of a process or a product in many, many cases.

John Higgins 10:39
So a lot of inspection companies develop protocols that are intense. And what they should really do is focus on those critical customer requirements. Very importantly, and not get lost into the detail that is, doesn't have a meaningful effect to the customer.

David Hoffmann 11:00
Right, Gianna, I get it. And you and I have had this discussion a lot. And I'd love to kind of go into it now. But because we've looked at how people inspect products at factories, and we say like the amount of testing and checking that's done, how necessary is most of it, because the current process, the current way of inspecting is based on AQL levels. So I'd like to go into AQL levels a little bit. And then I'd like to go into you know, what is the alternative to an AQL? level? And why might that be better? You know, if at all? So maybe just touch on kind of what exactly is an API level? And what how did even come about to be and why are we using it today?

John Higgins 11:42
Well, it started off in World War II, when they were manufacturing bullets for the US military. And as you can appreciate, you can't test all the bullets, otherwise, you have no one to use. So.

David Hoffmann 12:01
Make sense? Actually, yeah,

John Higgins 12:03
that's a destructive test. So you have to work on a sample. And that's what the famous quality experts of the time Dodge and Romig they came up with the AQL tables, which allowed one to select samples from the production lot and plot the sample with a confidence level. So that they didn't have to test all the all the all the products otherwise explained. In terms of the bullets, that would be 90 years. That's another eight kills were developed. And they develop tables based on attributes. And also based on variables. So there's two dimensions to the Ito tables, its attributes and variables, people are very familiar with the attribute one but not so familiar with the variable.

David Hoffmann 12:58
Okay, well, what what do you mean by attributes in a variable, John, if you can just explain to us

John Higgins 13:03
is like, is the product that k is it doesn't appear to perform well. So it's a descriptive test. The variable this is where you measure something like the length of a lead or the dimension of the diameter of a hole, or the weight of something or the voltage or the wattage required, right? Such

David Hoffmann 13:31
something that could vary and it's a very quick vary and may not be 100% consistent or replicable in every single item.

John Higgins 13:40
So there's two dimensions to AQL. And these and these two aims to AQL is the upper end, which is the the AQL. And the lower end, which is called tillable name Naima, LTPD lot tolerance percent defective. But the short term of that is reject quality levels at the one end of the AQL curve, you've got the reject quality level. And at the other end of the curve, you've got the AQL and accuracy provides a percentage of what you're trying to achieve in terms of good product. So, if you run IQ 1% That means that on average, in terms of your inspection, you provide a 95% confidence that the 1% of the products are likely to be defective or less right, wrong. Or QL is on the other side, which very few people pay attention to which is the bed in of the product. What are you What is the worst case that you can send to the customer through inspection? Because right as I said, the our confidence levels in the inspection because you're working on sample Yes.

David Hoffmann 15:03
Yeah. So it's kind of like taking a representative sample size and saying, like, what the upper limits and lower limits are and then say, well, based on this small sample size I've checked, we assuming that the rest is going to be the same. Right? And John, why? Why has a QL kind of persisted? I mean, everybody talks about AQL. When AQL level you're using, is it the most effective way of inspecting visit a factory these days? Are there alternative methodologies? I know, we've spoken in depth about the 50 piece representative sampling or homogeneous batch. Let's Can we talk around that a little bit, because I tend to agree with you and your thoughts on that, and AQL shouldn't necessarily be applied anymore. And the 50 piece homogeneous batch is more reliable and more effective. But you know, maybe you can break it down for us a little bit.

John Higgins 15:59
Okay, so, you know, inspection, you can inspect quality into a product, we have to build it. Yes. So there are a lot of dimensions in terms of crypto customer requirements. The approach should be to make the inspection is as effective and as quick as possible. Right, the way to achieve that is to get get close to the supplier, Ryan moleskin, you know, you're producing batches, keep the batches that are homogeneous together, so that we can build the homogeneous batch and not a heterogeneous batch. Because if we do a homogeneous batch, we can reduce our sample size. But if you're dealing with a heterogeneous batch, then your sample size will be based on the AQL tables as provided,

David Hoffmann 16:56
John, just for the sake of everybody listening, what is a homogeneous batch? And what exactly does that mean? I know what it means, but I'm not sure everybody else does.

John Higgins 17:09
homogeneous means the product has been produced with the same material, the same people, same equipment, same production line, same management, right. So the homogeneous means, therefore the product in the batch is similar in those those respects,

David Hoffmann 17:31
to the whole. So John, can I can I just, it means the whole batch that's made, is made the same way. Same, nothing has changed, right? With a minute you add any change that maybe you produce it the next day, maybe you continue producing the next day, and you even change workers or maybe get a new batch of material in, it's no longer a homogeneous batch, there's, there's a space that something could change or deviate or be different from the first production, right?

John Higgins 18:02
That's absolutely correct. So so to make the inspection effective, it is important to establish whether you're looking at a homogeneous batch to inspect or batch, if it's a big batch, then you're tied down to the AQL sampling plans as provided. But if you try to get it to a homogeneous batch, then there's opportunity to reduce the amount of inspection and achieve the same result as it was a heterogeneous pitch.

David Hoffmann 18:37
Right? So you've always said to me that 50 pieces from a homogeneous batch is kind of the best result possible. Is that true? And why do you say that and what is the science or, or or statistics behind it?

John Higgins 18:59
Okay, so if you look at the population, a batch of population homogeneous population, the if you look at the statistics around that, they talk about the Zed curve, okay? And if you take a song, a representative sample, from that, you can apply what they call the t test. Okay, which is a soulful test. But the statistically it's been proven that if the sample batch is equal to or greater than 30, it's representative of the population, the Zed population, and

David Hoffmann 19:39
CityJet y 50. John, that's what I've always been interested. I mean, I know what AQL if you've produced 1000, or 2000, or 3000, the table tells you pulled fossil, the 20 songs are very solid based on how you produced but you're now saying, as long as it's homogeneous 50s The number always is that true?

John Higgins 20:04
30 or more minimum is that that has been statistically proven, it's been statistically proven that a sample of 30 or more represents the population confidently. Okay. So without getting too statistical, that is the theory behind it.

David Hoffmann 20:27
Okay. As long as you can validate its homogeneous batch.

John Higgins 20:32
That's very important. Okay. Yeah. So if you do that, then you can apply this thinking immediately to your inspection. Right?

David Hoffmann 20:44
So would it be safe to say if you know your manufacturers producing batches each day that you could apply, save, inspect 50 per day from each production batch? Would that be reasonable? Or

John Higgins 21:00
100%? Correct? You think he is perfect?

David Hoffmann 21:03
Yeah, you know, I've long had this kind of thought process from just real life field experience of the last 20 years. We're looking at the level of inspection that goes into product sometimes, and I go, like, I think you hit the nail on the head, you can't inspect quality into a product, it's got to be built with quality already, the inspections just to pick up the potential problems that might have gone wrong in manufacturing, and making sure it's fit for market and find out the detail that goes into inspection. Sometimes it's so overkill, that sometimes obvious things are missed and reducing that sample size. But really going deep into the checking seems to make a lot of sense to me.

John Higgins 21:45
There's a requirement obviously, that the something must be random, that must not be biassed. So it must be a random sample. sample from a part of the batch, you treat the batches as one big batch, and you take all aspects of the batch is no bias in the sampling. That's really important.

David Hoffmann 22:07
Yeah. And John, what what are your thoughts or feelings in terms of, let's say you find defects or problems on a batch, and it's already been produced, packaged and ready for delivery? How do you weigh up or assess? Will eree Work meaning unpacking re handling, setting up production process to fix something more things can go wrong? With What's your kind of decision making factor to say? Well, it's it's less risk to ship it as is based on the problems found a verse hires to rework it may cause a whole bunch of other defects, or miss handling or even cosmetic defects, potentially, by reworking and how do you? What are your kind of rules of thumb in gauging them? I mean, there's some obvious answers, but I just wanted this approach in your language and world.

John Higgins 23:04
Okay, so the defect could be critical. If it's critical, then there's no discussion. If it refers to safety, then the whole batch must be reworked. Question

David Hoffmann 23:20
is safety, the only benchmark for critical or the other one of the top chemo

John Higgins 23:25
could could potentially cause a very expensive repair. That could also be a critical requirement. Run, it depends if it's now a minor. If it's a minor, one one could say to the customer, could you provide us a concession based on the amount of funding and we'll leave that to the customer to make a final decision? If the customer says no, in killing, you got to rework it. And a major would also be consideration similar to net. Vijay is in between the critical and the minor? So I would suggest that if you have a critical, there's no discussion, you don't even go back to your customer. You instituted complete rework and address that problem without discussion. But if it's a major or minor outage, engage the customer to on whether this should be reworked or not.

David Hoffmann 24:26
Yeah, well, there was always two sides to that, right, because everybody wants to deliver a perfect product. And it's, I guess, it comes down to the way I look at is what would a consumer accept or not accept and what is the cost of the consumer not accepting that? So I try to boil down to common sense. What I find John and this is always an ongoing challenge is very few people define upfront, what they classify as major, minor or critical in terms of What they're going to accept as a finished product? I mean, can you speak to that a little bit? I mean, how does a regular person is purchasing products, for example, out of China, establish a list of majors, minors and criticals to share with the factory upfront.

John Higgins 25:18
The inspection company should have qualified engineers that could interpret the impact of the defect finding, and then assist the customer in making a decision. But what we must also be very careful of is that the minute the manufacturer, the supplier, doesn't entrap the inspection company into doing the inspection for them, they must have, they must do their own inspection, according to the quality management System ISO 9000, perfectly, and not in an inspection company to be the inspection, if you know what I mean. Yeah, 100%

David Hoffmann 26:04
Yeah, that they needed inspector and possibly, you know, I always say we should check what their quality levels are internally first, and how they post their own goods. And if we in agreement with that, we just stuck a second check out of it.

John Higgins 26:19
So what our do the inspection company, if they find a defect, I would say they should go to the the suppliers, quality inspections and ask if they found such a defect. And if there's some correlation, then we know we got a serious problem, but it is no correlation. It could be just an outlier that occurred by chance. So these are the considerations that one should take into place into mind.

David Hoffmann 26:49
Yeah, that makes a lot of sense. And when you almost ideally want to be in a situation with a factory, who's got their own quality control systems, and can do it without your need for inspection, in theory, but and then your inspection, just kind of your third party safety net. Yeah, I get that. And that's often misunderstood. And it comes back to that whole thing, which I live by what you said a long time ago, to me, you can expect quality into a product, it's got to be inherent in the design and the way it's built in the way it's made. And so on and so on.

John Higgins 27:26
So there's another consideration. Now, when you're doing inspection, you're switching around and you're you're looking and testing a product for maybe five or 10 minutes maximum. But what what happens if the failure as a time dimension to it and it fails? After a couple of weeks or months? Yes, you're not likely to pick that up in your in your inspection. So how are you addressing it? Well, it's, it's really simple. If the supplier who's providing the product is not doing reliability tests, then that's a concern. So one should always study the reliability testing to make sure that there's no hidden defects that are time based. And company can do as part of the service to the customer, is to say, well, we will study the reliability testing, some of them good and heat chambers, some good ambient environment, it just depends of the way they structured the reliability testing.

David Hoffmann 28:35

Yeah, that makes. Yeah, that's critical. And sometimes also, reliability tests are done. Coming back to the homogeneous concept on products that were produced at different times to your production side doesn't mean your production matches the original testing they did you got to keep on checking that either by lab or internally, they have the facilities. Yeah, it makes a lot of sense. John, I know this is a lot for people to to digest. So I'm probably going to do some infographics and some nice help gods for the wind, we released this podcast. But what I wanted to close off by is just sharing. You just reminded me of a video that you once showed me and I'm going to put that video in this podcast so that people can watch it because it's so fascinating when we were talking about improvement processes in a business or factory production. Are you talking about how how improvements in a process speed things up? And I'll never forget the video you showed me. I've still got an edit to this podcast of the Formula One in 1950 or 60 or something I forget. And they did that when the car comes into the pitstop and they've got all the people doing the tire changes and do cleaner whiskey wipe and it's like, literally like a fart. I don't know maybe a two minute process. So as and then, you know, the car drives off. And then they fast forward to like 2016. Formula and car comes in and they knock under two seconds, they've changed every every time the windscreen the cars have. And it's the perfect way to understand how, like, processes change and can be super efficient and more effective. And like that blew my mind. We just saw how they were doing the work before and into how it became better. But it really puts into context, the fact that you can't ever say anything's wrong. It's just how can I make a better faster, quicker? I mean, and it's from two minutes to one second, it still blows my mind. But you would say it's impossible, yet it's not it's