May 28 2015
Why Your Lean and Six Sigma Improvement Efforts Aren’t Driving Better Results | IndustryWeek | John Dyer
“Don’t expect a positive ROI from your lean and Six Sigma investments if they are nothing but a pretty picture.
I once had a plant manager tell me his factory had implemented Six Sigma, but there was not a single statistical process control chart. How is that possible? Another had the control charts in place but refused to allow the operator to shut the process down when it indicated an out-of-control condition. Another plant claimed it was lean but had a dozen bins of parts stacked on the floor as part of a two bin system. Another plant routinely violated the daily production plan by rescheduling orders, and then the plant blamed the supply chain for causing it to frequently run out of parts (which then drove it to change the schedule… a vicious circle).”
Source: www.industryweek.com
Michel Baudin‘s comments:
A good article about how you fail to achieve any performance improvement from what you pretend to do. The pretense doesn’t make you any better at what you do, but it may have a business purpose in getting you the opportunity to do it at all.
The author opens with SPC control charts, which aren’t even part of Six Sigma as originally developed. These control charts are a 90-year-old tool that is not needed in mature industries and cannot cut the analytical mustard in high technology. It’s a “Three Sigma” tool that Six Sigma was supposed to replace, until it was watered down to the old SPC.
Is posting SPC control charts on your shop floor useful? Not if you want to improve your process capability, but yes if you have a customer who demands to see them. The tangible result is the contract you sign with this customer.
Many manufacturing organizations also pretend to “do Lean” to humor customers. It’s a cost of doing business. What you do for show does not, and cannot improve performance because it’s not the intent.
Mark Graban called this L.A.M.E. for “Lean As Mistakenly Executed”; Bob Emiliani, fake Lean. I just call it “Lean Lite.” Those who really want to improve performance are annoyed by it, because it distracts manufacturing professionals and, over time, devalues the “Lean” label.
If your focus is performance improvement and not appearances, you select projects that have an impact and develop your skills — like SMED on an injection molding machine — but may not be as visible to visitors. You get more checkmarks on their lists by hanging Andon lights on all the machines, even if no one pays any attention to them in daily operations.
If you are an engineer, all you care about is making things work. In marketing, on the other hand, appearances are essential. To be a general manager, you must be able to see the business from both perspectives.
See on Scoop.it – lean manufacturing
May 28, 2015 @ 7:14 am
Well, I too would criticise the author. Why would you shut a process down just because of a single out of control observation?
But, if process behaviour charts aren’t part of Six Sigma it’s only because it’s assumed you’re already doing them. What else do you use to manage statistical control/ stability and predictability/ exchangeability? I’m happy to listen to suggestions.
May 28, 2015 @ 9:50 am
I am not sure what you mean exactly by “process behaviour charts.” I was referring specifically to control charts, as described in the literature on Statistical Process Control.
Statistical methods for improving process capability are useful when your processes are not capable. In mature industries, like automotive, you have machines that can hold tolerances tens times narrower than required, and process capability is not an issue there. Quality problems in such processes are caused by discrete events, like tools breaking, and human error. One-piece flow lets you detect problems quickly, and you organize for rapid response with methods like QRQC.
Where you have process capability issues is in industries like semiconductors where, if your processes are mature, your product is obsolete. There, you need statistical tools, but control charts are too primitive for this purpose. Instead of a trickle of manually collected,one-dimensional data, you have a flood of automatically acquired multi-dimensional data. You need advanced data mining tools for the data you routinely collect, and statistical design of experiments to improve your processes.
May 29, 2015 @ 2:35 pm
Michel, to be a general manager, would you also add the perspective of the customer, or is that just a given? Or is it ultimately a combination of engineering and marketing?
For myself, I often see the customer’s perspective forgotten in favour of countermeaures that appeal more to the person implementing or approving them.
May 30, 2015 @ 11:36 am
Comment in the IndustryWeek Manufacturing Network discussion group on LindedIn:
May 30, 2015 @ 11:39 am
I agree that it would be great if every employee had a basic understanding of probability and statistics. However, in spite of advocacy for the past 85 years by Shewhart, Deming, Mikel Harry, and others, it hasn’t happened, and I doubt that it will over the next 85 years.
To be effective at establishing or improving process capability, you need to understand both the physics or chemistry of the process, and the art of analyzing data. You can muddle through with knowing only the process, and occasionally only with data science, but you need both to fire on all cylinders. The problem you run into is the scarcity of people with both sets of skills, and the point of the black belts system in Six Sigma was to grow more of them. This challenge, incidentally, is not particular to manufacturing.
You single out the Normal distribution as a concept everybody should understand. I prefer to call it Gaussian, because calling it “normal” leads people to assume that every other distribution is, in some way “abnormal.”
May 31, 2015 @ 7:05 am
Comment in the IndustryWeek Manufacturing Network discussion group on LindedIn:
May 31, 2015 @ 7:09 am
My experience is consistent with yours. I am going to date myself here, but, when I started out as an engineer, QC circles were in fashion in the US. I was trained as a QC circle leader and discovered that operators took to the “7 tools of QC” rather well. Within a year, however, there was a management change, and the new VP of manufacturing got rid of all this statistical mumbo-jumbo and went instead for the Crosby approach, as a result of which the quality improvement effort boiled down to the slogan “Do it right the first time!”
Later, when I had kids, I discovered that the same statistical concepts were taught in American middle schools, with 7th and 8th-graders learning about correlation and regression and working through reasonably sophisticated examples.
So there is no doubt that it is teachable, and, according to your post, you have been able to reach 4000 of the 11 million people working in manufacturing in the US today. The problem is not with the operators’ real attitudes and cognitive abilities, but with managers’ perceptions and an ideology that still make reward and punish based on luck, as in Deming’s red bead game.
Regarding your Process Improvement Engineers (PIE), I was wondering how much authority you give them. Whether they are called Production Supervisor, Group Coordinator, …, I found that the first-line managers are the key agents of change on the shop floor. Upper management commonly underestimates their potential and keeps their numbers too low. You often have 1 for every 100 operators, who can do little more than take attendance, enforce discipline, and expedite parts, when you really should have 1 for about 20 operators, working on improvement and people development. It is a broader issue than statistics, but it brings us back to the original topic of this discussion.
I recommended to a general manager that he increase the number of supervisors by giving some of his numerous engineers the opportunity to take on this role, which several eagerly accepted. This put people very much like your PIEs directly in charge of operations, and it did more to improve performance than just about anything else. Some of these engineers-turned-supervisors later returned to Engineering, but others stayed, and one of them later became plant manager.
May 31, 2015 @ 8:23 am
Michel – response to your 28 May post – I have adopted Don Wheeler’s suggestion of using “process behaviour chart” (PBC) over “control chart” as it’s just a better description. I also prefer his “stable and predictable” over “in statistical control”.
I agree that if you have Cpk in the double digits and no threats disclosed by a rigorous FMEA then you probably have better things to do than plot PBCs routinely. However, PBCs are exactly the right tool for working to improve human error and tool breakages in the same way that they were for issues of mere dimensional precision.
Further, if you have the ambition to embark on designed experiments then maintaining stability throughout the experiment, by means of PBCs, is essential. Data mining is pure folly unless on the basis of a stable and predictable process.