Today I submit a quick, but extremely important post if you produce a product of any kind.

There are two different ways your product can be “bad.”

It’s defective

First, said product can be defective. This is to say something is wrong with it. It doesn’t work like it’s supposed to. It doesn’t look like it’s supposed to. The customer, who pays you good money, ain’t gonna be happy with said product.

It has multiple defects

Second, the aforementioned defective product might have multiple defects. This is to say there are many things ‘wrong’ with the product. It has 103 different scratches all over it, 14 surface mount components are not soldered correctly, and the text on the label is incorrect in 3 different places.

Manage the defects

In summary, it only takes one defect to create a defective product. But a single defective product can have a zillion defects.

What does this me to you, maker of said product? You need to stop focusing on how many defective units you produce… instead you should be laser focused on the number of defects you’ve created the past month.

Do you agree? Is this how you manage your business?

As promised a few posts ago, I’d like to explain exactly how to go about completing a measurement systems analysis (MSA).

If the sound of this makes you want to yawn… well… stop it! It’s Monday and time to get focused (and some coffee).   Seriously, this one tool has greatly benefited me over the years and I know it will you too.

The two pre-cursor articles to this one explained one such example of a variation story and how a simple MSA helped save millions of dollars in 60 minutes.

Now then, tonight I want to begin by explaining what a measurement system analysis is as well as two types of MSA that will serve you very well once you’ve mastered them.

MSA Defined (Job Interview Proof)

Right, so if you’re ever on a job interview and get asked to explain what an MSA is you would probably do well to rattle off something like this:

A Measurement System Analysis, abbreviated MSA, is a specially designed experiment that seeks to identify the components of variation in the measurement.

That would be the Wikipedia definition and I guess it sounds pretty nice… and chances are the guy interviewing won’t have a clue himself… so, yeah, this would be a great answer!

MSA Defined (Ron’s Take)

But listen, I’m a simple man. So rather than getting all text book like allow me to offer my explanation of what a MSA is. You see, anytime we measure something (e.g. the weight of a part) we’re actually dealing with two types of variation.

1. The variation within the parts/process we’re studying (P)
2. The variation within the measurement system itself (M)

In other words, the total variation we’re experiencing (T) can be explained by a formula that looks something like this: T = P + M.

In a perfect world the variation due to M (measurement system) would be zero meaning all we need to do is attack and defeat the variation within the process (P) and then ride off victorious.

Sadly, many times there is so much variation with the actual measurement system we have no earthly clue how much variation is due to the actual process.  Even worse, we may try and try and try to improve the process only to find out the problem wasn’t even the process… it was the dang measurement system!

And this is precisely why measurement systems analysis is so important. We must identify how much of the total variation is due to the process and how much is due to the actual measurement system.

Types of MSA

There are different types of MSA out there and the one you choose has a lot to do with the data you’re dealing with.

• When dealing with variable data (data we can measure like weight, distance, etc.) a Gage R&R study will serve you well. This stands for Gage Repeatability & Reproducibility.
• When dealing with attribute data (pass/fail, good/bad, etc.) an Attribute Gage study will do the trick nicely.

Up Next

Now that we know what a MSA is as well as two popular types of analysis it’s time to learn how to conduct the actual studies… and that’s precisely what we’ll discuss in the very near future!

If you have any experience, good or bad, with measurement systems analysis please share in the comments section below.  Also, if you have a specific question on the topic please share that as well.

Last night I told you the story of a measurement system analysis gone bad. Tonight, as promised, I’ll explain exactly what the issue was and how we solved it in around 60 minutes.

The Part in Question

The plastic component in question measured approximately 4” by 2”. I’ve made a rough sketch of what it looked like (see picture).

The component was actually completely black in color but I’ve made it gray in order to point out some key characteristics.

Towards the corners of the part were two key features.

First, points “A” were a rounded black protrusion-like piece of plastic that stood up about 1/16 of an inch in three corners of the piece.

This is what the camera looked at when taking the actual measurements. The camera would first “zero itself” and then it would measure the distance of the three black protrusions from one other. From this the machine would calculate the X and Y dimensions of the part.

Next, points “B” were drilled out circles that were used the hold the part into the final product during assembly. These circles were ignored during the measurement process.

30 Minutes of Brainstorming

After we failed the MSA we all sat around a table and talked about what the issue could be. We felt sure that the actual measurement system was probably OK and that the problem had more to do with the measuring process itself.

So we started throwing out ideas… everything from the fixtures we used to the way the operator loaded the parts was discussed.

Then someone wondered if looking at the black protrusions was the best approach. This person’s theory was that the camera may be struggling to distinguish the difference between the black protrusion and the black frame.

After a few more minutes of pondering the idea of using the circles right next to the black protrusions was tossed out there.

The supplier’s engineer was skeptical. He didn’t think it would work. But after some old fashioned Texas hog wrestling (a figure of speech, no actual wrestling occurred) everyone agreed to give it a shot.

30 Minutes of Trystorming

Using the same 15 pieces and the same operators and a modified computer program we re-did the MSA in about 30 minutes. The difference this time was we used the drilled out circles instead of the black protrusions.

Once we had the data we re-did the analysis in our statistical software package. The result? We passed the MSA with flying colors.

Problem Solved

And that was that. Some simple experimentation coupled with both brainstorming and then trystorming solved a problem that had been plaguing this company for the better part of a year.  The best thing was we started in the morning and were done by lunch time.

We were also able to proceed with our DOE and optimize the injection molding process. Soon, the quality problems we were experiencing from this supplier completely vanished.  It was truly a win/win situation.

The key learning for me was that continuous improvement doesn’t have to take a long time. Problems can, and in my opinion should, be attacked and solved quickly.

It’s Time to Learn How to Conduct an MSA

Now then, in light of this story I thought it may be beneficial for me to explain exactly how to go about doing a measurement systems analysis. And that’s precisely what we’ll do over the next few posts.

So stay tuned and please forward this post to anyone who may benefit from this type of information by clicking the “Share This” button below.  From there you can email the article.

If you test or inspect (visual or automatic) a product of any kind I’m very confident I can show you how to save your company thousands, if not millions, of dollars (or Euros, Pounds, etc.).

How you ask? Three words. Measurement Systems Analysis, or MSA for short. But first a true story.

Back in the Day

In a past life I was working with a supplier that provided my former company with some critical plastic components for the product we produced. I’m purposely being vague to protect the innocent!

Anyhow, we were having some problems with the aforementioned critical components so I went to the supplier to see if I could help.

After some investigation and discussion we decided to run a design of experiment (DOE) on the injection molding process that produced these parts.

A Simple Question

Being the good six sigma guy I try to be, the first thing I asked the engineer I was working with was if they had ever done a measurement systems analysis of the optical inspection machine used to measure the actual dimensions of the parts.

He went on to explain how this very expensive inspection machine had just recently been calibrated to the “gold standard.” I wasn’t quite sure what the gold standard was… nor did I care how calibrated it was.

I asked him again, “Have you ever done an MSA.” He was growing frustrated with me and finally admitted… no, they had never done an MSA.

Let’s do it!

Happy to finally hear this I asked him to collect 15 of the parts we were planning to measure.

About 8 minutes later he had the 15 parts.

We grabbed a couple of operators and proceeded to conduct a variable data measurement systems analysis. It took us about 30 minutes to complete the entire study.

I threw the data into some statistical software and did the analysis.

Houston, we have a problem

Guess what? We failed the test. And I don’t mean fail like we did sort of bad. I mean fail like this fancy optical inspection machine that had recently been calibrated to the gold standard couldn’t tell the difference between a good part and bad part if its life depended on it. Good thing it wasn’t alive!

In other words, this supplier had no earthly clue what the *true* dimensions of these components were.  Not good.

The Rest of the Story

Tomorrow night I’ll explain exactly how we solved the problem in less than 60 minutes. By the way, the cost of poor quality of this particular problem was resulting in millions of dollars of damage each year.