iSixSigma

Reliability at TOYOTA. How far is 6 sigma quality?

Six Sigma – iSixSigma Forums Old Forums General Reliability at TOYOTA. How far is 6 sigma quality?

Viewing 56 posts - 1 through 56 (of 56 total)
  • Author
    Posts
  • #36051

    REDFSS
    Participant

    Recently published figures in one of the leading newspaper showed the reliability scores of Toyota, American Honda and Porsche. Unit of measurement = Defects per 100 vehicles.
    1. Toyota was leading with 207 defects / 100 vehicles.
    2. Next was American Honda with 210 defect / 100 vehicles.
    3. Porsche had if I remember correctly 237 defects / 100 vehicles.
    Thats all fine. But there is no information about the aver. opportunities per vehicle. Has anybody got an idea. That is certainly not a true Reliability measure.
    If we assume the following opportunities, can we say still that no one among them is still near Six Sigma? Please correct my DPMO figures if you have them near to correct ones. They are just a wild guess.

     
    Defects
    Vehicles
    DPU
    Opp. Per unit
     DPMO

                 Zlt  
             Zst

    Toyota
    207
    100
    2.07
    200.00
    0.01035

    2.31
    3.81

    Honda
    210
    100
    2.10
    200.00
    0.01050

     2.30
    3.80

    Porsche
    237
    100
    2.37
    150.00
    0.01580

    2.14
    3.64
    Thanks,
    REDFSS

    0
    #102819

    REDFSS 2
    Participant

    Sorry. Earlier post had DPMO instead of DPO. DPMO column was missing.
    Thanks,
    REDFSS

     
    Defects
    Vehicles
    DPU
    Opp. Per unit
    DPO
    DPMO
    Zlt
    Zst

    Toyota
    207
    100
    2.07
    200.00
    0.01035
    10350.00
    2.31
    3.81

    Honda
    210
    100
    2.10
    200.00
    0.01050
    10500.00
    2.31
    3.81

    Porsche
    237
    100
    2.37
    150.00
    0.01580
    15800.00
    2.15
    3.65

    0
    #102850

    why,oh,why?
    Member

    This reminds me of a recent thread when a certain author/consultant took his magic wand and tried to assign a sigma value to NASA…
    Please take no offense, but ‘Yer in fer a heap of trouble here, partner’  …..Are you with me, regulars? FYI, regarding this thread, look back in the search to earlier this spring using NASA, and Scorecard, if memory serves. Good luck.

    0
    #102881

    V. Laxmanan
    Member

    Dear RedFss:

    Defects
    Vehicles
    DPU
    Opp. Per unit
    DPO
    DPMO
    Zlt
    Zst

    Toyota
    207
    100
    2.07
    200.00
    0.01035
    10350.00
    2.31
    3.81

    Honda
    210
    100
    2.10
    200.00
    0.01050
    10500.00
    2.31
    3.81

    Porsche
    237
    100
    2.37
    150.00
    0.01580
    15800.00
    2.15
    3.65

    I notice something curious here.
    Take Honda, 210 = 70 x 3. For Porsche, 237 = 3 x 79 and for Toyota 207 = (3 x 70) – 3. 
    In other words, if x is vehicles and y is defects, we get y = hx + c. The constant h = 3 for all three brands and the constant c = 0 for Honda and Porsche while c = – 3 for Toyota.  We can use this logic and look at the defects data for all the vehicles in the J. D. Power surveys.
    Unfortunately, I do not know how many vehicles were actually tested. Was it exactly 100 for each brand?  Did the number x vary? I have not been able to get this info from published sources.
    There are many different ways of distributing a fixed number of defects beween a fixed number of vehicles (say 100). So, I tried to “simulate” the evolution of 237 defects in a survey with 100 vehicles.  I will post these results separately. 
    The pressure of a gas is the external manifestation of the force exerted by zillion of microscopic entities (called atoms and molecules) that are moving about chaotically and collide with each other and the walls of the container.  In the 18th century, Daniel Bernoulli was the first to apply Newton’s laws of mechanics to the particles of a gas and should that the inverse relation between pressure p and volume V (known as Boyle’s law) follows from this “model” of what is going on in a “gas”. 
    The next steps were taken by Rudolf Clausius (who conceived the idea of “entropy”) and by James Clerk Maxwell, more than 100 years later, in the second half of the 19th century.  Bernoulli’s contribution was largely forgotten.  Maxwell applied the laws of probability and statistics (he uses the basic Gaussian distribution) to solve this problem, and arrives at a very elegant formula for the distribution of molecular velocities.  But there were some difficulties with the theory since it did not agree with experimental observations.  Essentially, the theory could not account for the ratio of the specific heats of gases, heated at constant pressure (holding V constant) and at constant volume (holding pressure p constant). Maxwell was gratly disappointed. This difficulty was finally resolved after Einstein applied Planck idea of a quantum of energy to explain the behavior of another macroscopic property called the specific heat of a body and how it varies with temperature. 
    Likewise, the defects per vehicle, or defects per 100 vehicles is also an external manifestation of what is going within a large organization like Toyotal, or Honda, or Porsche. Thanks for calling this to the attention of the forum.  Regards.
    Laxman 

    0
    #102883

    New BB
    Participant

    I too computed the sigma level for the posted automobile
    data and got the same results as REDFSS.

    0
    #102884

    V. Laxmanan
    Member

    Dear New BB:
    You are absolutely right. I did that too. 
    If you make some assumptions about opportunities to produce a  defect per vehicle, which then gives the estimates for DPO and DPMO, you will get the Process Sigma, with and without the 1.5 shift given in the last two columns of the RedFss post. 
    I am trying to do something different now.  I just finished that and will said to the forum for downloading. If you are interested, do send me an email at [email protected] and I would be happy to email the files to you separately, so you take a look at it before it is posted here. Thanks a lot. Regards.
    Laxman

    0
    #102917

    REDFSS
    Participant

    Sorry friend I am just curious to analyze the data from the Reliability information published and not trying to attempt anything with crazy assumptions and I am not interested in talking about NASA. No!
    To me it seems quite a good topic, which can further unveil valuable information on other Reliability and Six Sigma metrics. I do not want to go with only assumptions and hypothetical CTQs like NASA thread.
    Questions are simple. Based on the reliability data published can anybody help me getting close to the process sigma capability of these automobile producers? Are the reliability measures correct? Maybe, internally they have MTBF, Defect Arrival Patterns and Reliability Growth studies conducted. How do I arrive at correct DPO values? Where can we find the data published?
    I do not want to assign a sigma value to TOYOTA or others. The data comes directly from the customers I presume. Secondly, if we say there are “defects” per 100 vehicles then there must be several opportunities during testing. Do we have this data published from each producer? Or anyone from these three companies can give us the information? I do not see anything wrong in providing forum with this data. At least close numbers.
    And I do not want to go and get entangled in the threads on NASA Six Sigma. It will take a month looking at my current tasks and available time.
    Thanks.
    REDFSS

    0
    #102918

    V. Laxmanan
    Member

    Dear RedFss:
    Questions are simple. Based on the reliability data published can anybody help me getting close to the process sigma capability of these automobile producers? Are the reliability measures correct? ?
    I do not want to assign a sigma value to TOYOTA or others.  I do not see anything wrong in providing forum with this data. At least close numbers…
    If you have “close” numbers, or some data, please post them and let’s take a look at it. Regards.
    Laxman

    0
    #102919

    REDFSS
    Participant

    Laxman,
    I think its a communication error. I meant I do not have the numbers that could be considered “close” to the “actual” ones available with these companies.
    If we have these “true” numbers, nothing worth it !Isn’t it.
    But I feel someone having worked on all these or anyone of these automobiles could be of great help.
    Cheers !
    Redfss

    0
    #102920

    V. Laxmanan
    Member

    Dear RedFss:
    I thought you were trying to make sense of some of the numbers you have (about automotive industry process capabilities) with the estimates that were trying to make using the defects per vehicle from the J. D. Power Reliability survey. 
    I agree with your general premise – that may be we can learn something about what is going on within these companies and their process capabilities by analyzing the J. D. Power data. From a model for the microscopic behavior, we can arrive at macroscopic behavior.  Or we can infer something about the microscale behavior from the macroscale observations (in this case Defects per 100 vehicles). 
    I have sent my documents for downloading to the forum, where I tried to analyze the J. D. Power data, at the “microscopic” level, prompted mainly by your post. It should be available, hopefully today. Regards.
    Laxman

    0
    #102923

    Gabriel
    Participant

    You must be kidding:
    “Toyota 207 = (3 x 70) – 3”
    What about Toyota 207 = 3 x 69, and we forget about the -3?
    Also:
    “Honda 210 = 3 x 70”
    Why not Honda 210 = 3 x 69 – 3?
    And for Porsche we can keep the 70 and the h = 3 too:
    Instead of 237 = 3 x 79, we can say 237 = 3 x 70 + 27.
    Of course you can also say 207 = 9 x 23, 210 = 9 x 23 + 3, and 237 = 9 x 23 + 30 or = 9 x 27 – 6 or  = 9 x 26 +  3 or….. And what would that mean?
    What about 207 = 2.7 x 100, 210 = 2.1 x 100, and 237 = 2.37 x 100?
    Sorry, but I don’t get to understand the logic you use to “choose” the values of h, x and c (among the infinite combinations available) you put in you linear equations. They seem to be put by the power of your finger, rather than by the nature or behaviour that the data shows.

    0
    #102925

    V. Laxmanan
    Member

    Dear Gabriel:
    Sorry, but I don’t get to understand the logic you use to “choose” the values of h, x and c (among the infinite combinations available) you put in you linear equations. They seem to be put by the power of your finger, rather than by the nature or behaviour that the data shows.
    That’s exactly what I am trying to explain.  There is a “nature of behavior” of data that I see that other do not seem to see. I have sent some documents for downloading to the forum which discusses this in more detail.  They should be posted later today.
    Ultimately, everyone has to think about the work function idea and come to grips with it. I do realize that there is only so much I can do here. Have a nice day. With my best regards.
    Laxman

    0
    #102928

    Gabriel
    Participant

    But you can help me with that. You showed us three equations:
    Honda: 210 = 70 x 3
    Porsche, 237 = 3 x 79
    Toyota 207 = (3 x 70) – 3
    I can only see where the 210, 237 and 207 come from (it is the ration defects / 100 vehicles, as they were posted).
    Why don’t you just explain us (or to me at least, you have my e-mail if you want to do it off-line), in a simple and stright way, how did you get the 9 numbers at the right of the “=”? (I am aying 9 because I am counting 2 independet terms not shown that are zero). I mean the process you followed to arrive to that numbers, using this example and these numbers.

    0
    #102932

    V. Laxmanan
    Member

    Dear Gabriel:
    Thanks, I would be quite happy to.  In the interim I see that some the files sent to the forum are now posted. Please do study them.  For someone who I sincerely admire, based on what I seen in your posts, this should not be too difficult. But, …. 
    you have my e-mail if you want to do it off-line),
    I really do not have your email to do it off-line.  If you can provide it, I would be very happy to do exactly what you have suggested. Regards.
    Laxman

    0
    #102936

    Gabriel
    Participant

    Last Friday or Sutarday, I sent you an e-mail to the e-address you submited in one of your posts. Since I did not receive a rejection message from a server, I assumed that it had been successfully delivered. Can you check, please?

    0
    #102940

    Noname expert
    Participant

    A Real Surprise ?What about all arguments explaining the Toyota’s TPS and the SS efforts in Honda and GM?Somebody has to explain as we are totally confused????????

    0
    #102942

    Noname expert
    Participant

    Thank You.I was the one  who have raised the (old) question  about  SS Score at  NASA.I’m really happy that you link this  topic to the Nasa’s score as I’m still not satisfied with the previous answers,also I’m eager to hear a convinving answer here as well.thanks and regards.

    0
    #102943

    Gabriel
    Participant

    Dear V,
    I have just read the two documents. Something intrigues me.
    Let’s say that I peek some 100 cars to make the defects survey. Each car has a given number of defects that I don’t know forehand.
    However, any conclusion I derive from that survey must be independent of the order in which these vehicles ispected, because there is absolutely no “logical” or “preferred” order. They are not consecutive in the line of production, for example. So, as opposed to a control chart, the relation “# inspected” = “time” does not exist, so one cannot say “at this time we were producing this defect and ythen we fixed it”, or something like that.
    So, I propose that you do the following exercise: Take the data you got from the 100 “simulated” cars. Put them in 10 different random orders. Make the analysis. Do you get to the same conclusions? If not, your method is flawed. The conclusion can not be chance-dependant.

    0
    #102944

    V. Laxmanan
    Member

    Dear Gabriel:
      How nice to hear this from you! 
       Indeed, I just sent files to the forum for downloading, describing exactly what you have proposed.  They should be posted soon, I hope.
       Before I sent these additional files, I wanted to wait for some reaction and you have provided me the needed incentive today.  
       Instead of 10 simulations, I decided to do 4 simulations, holding the number of defects constant and 207 among the 100 vehicles. We can do 10, or for that matter even 20, 50, or 100 such simulations.  It would be much easier to ask J. D. Power to release the full data.  I am sure it is there, but nobody thought that was relevant.  If this creates interest, instead of doing simulations, I propose that we look at the “raw” data from this and other such surveys on hundreds of products whose quality and reliability are being tested.  As you can see h will change, if we want to make the automaker look good.  But, it is really the diferences in the work function that tell us the differences in the defect levels (at fixed x and at fixed h). 
      I have also addressed the question you posed with my arithmetical manipulations to arrive at h = 3 for 210 and 237 and how I then said the same h applies to Toyota as well.  I was thinking x = 70 for Toyota, or three less defects than for Honda, with x = 70 for both.  Note that the J. D. Power Survey gives the defect per 100 vehicles but not how many were tested.  Figures A, B, and C are meant to answer your question. If not, I have to try again.
      I checked my emails and I was unable to locate the email from you.  I would have been more than happy to respond to you – indeed honored to do so.  I did delete a lot of emails this weekend.  There were many in the inbox (not to mention many that I just delete with all the trash we get!).  Perhaps, I inadvertently deleted yours.  My mistake!  Here’s my email again, just in case, [email protected]
       Finally, as I have noted earlier, I am not a statistician by training.  I learned elementary statistics and have been doing a lot of reading on this subject lately.  For this effort to come to fruition, we need others -with solid statistics credentials – appreciating the value of the work function. Anyway, I am open for suggestions and learning still. Regards.
    Laxman

    0
    #102945

    V. Laxmanan
    Member

    Dear No name expert on Monday:
      I don’t know either about this TPS, TQM, SS, Lean SS, etc. I just took the numbers given on the defects and waved the magic wand with y = hx + c. Cheers!
    Laxman

    0
    #102947

    Noname expert
    Participant

    Thank you for your humble positive opinion,kind regards

    0
    #102961

    REDFSS
    Participant

    Why is the original question being hijacked and leading to just confusion?
    Why can’t we just stop drifting away with couple of people trying to convince each other in vain. Please keep your theory apart. Long long responses and those stuffed “pdfs”.
    Posting the questions again since I am forced to. No satisfactory answers I could really get, just discussion going everywhere on y = hx + c.
    ==========================================
    # 1. Recently published figures in one of the leading newspaper showed the reliability scores of Toyota, American Honda and Porsche. Unit of measurement = Defects per 100 vehicles.
    1. Toyota was leading with 207 defects / 100 vehicles.
    2. Next was American Honda with 210 defect / 100 vehicles.
    3. Porsche had if I remember correctly 237 defects / 100 vehicles.
    Thats all fine. But there is no information about the aver. opportunities per vehicle. Has anybody got an idea? That is certainly not a true Reliability measure.
    If we assume the following opportunities, can we say still that no one among them is still near Six Sigma? Please correct my DPMO figures if you have them near to correct ones. They are just a wild guess.

     

    Defects

    Vehicles

    DPU

    Opp. Per unit

     DPO

    DPMO

    Zlt  

    Zst

    Toyota

    207

    100

    2.07

    200.00

    0.01035

    10350

    2.31

    3.81

    Honda

    210

    100

    2.10

    200.00

    0.01050

    10500

    2.30

    3.80

    Porsche

    237

    100

    2.37

    150.00

    0.01580

    15800

    2.14

    3.64
    ==================================================
    # 2. I am just curious to analyze the data from the Reliability information published and not trying to attempt anything with crazy assumptions.
    To me it seems quite a good topic, which can further unveil valuable information on other Reliability and Six Sigma metrics. I do not want to go with only assumptions and hypothetical CTQs like NASA thread nor with someone’s pet – y = hx + c.
    Questions are simple. Based on the reliability data published can anybody help me getting close to the process sigma capability of these automobile producers? Are the reliability measures correct? Maybe, they use this in calculating MTBF or Defect Arrival Pattern or Reliability Growth studies. How do I arrive at correct DPO values? Where can we find the data published?
    I do not want to assign a sigma value to TOYOTA or others. The data comes directly from customer survey I presume. Secondly, if we say there are “defects” per 100 vehicles then there must be several opportunities during testing or complaint resolution. What are these opportunities? What is the correct number of opportunity? Do we have this data published from each producer? Or anyone from these three companies can give us the information?
    Thanks.
    REDFSS

    0
    #102963

    Mikel
    Member

    The simple answer is no – you can’t get the process sigma from the reliability numbers.

    0
    #102966

    REDFSS
    Participant

    Thanks Stan.
    But you have not provided any satisfactory answer. Maybe, you missed all the questions.
     
     

    0
    #102967

    Peppe
    Participant

    I believe to answer to your question correctly, we need :
    Bill of materiel for each type of car of each manufacturer (I believe is always the same for every manufacturer) to the parts that composed the car in the way to know the DPO.
    The life time for each car (could be the same)
    Another parameters are the price and level of electronics equipped , in the way to rank all performances at same level.
    But,  frankly speaking, with these few data available, I believe we cannot calculate anyway in right way, but only calculate every car as one opportunity for defect and made very simple (unreliable) calculation of  defect levels (as reported in previous posts).
    Maybe, here, anyone have that information available, otherwise they will be shared.
    Rgs,
    Peppe

    0
    #102969

    V. Laxmanan
    Member

    Dear Peppe, RedFss, All:
      Yes, we don’t know many things – not even how many vehicles were actually tested! 
      May be they did test 100 vehicles exactly, but I don’t know for sure.
      I was trying to show how to arrive at an estimate of Process Sigma, which was the original intent of RedFss, taking a different route. RefFss had to make some assumptions about how many opportunities to produce a defect per vehicle tested. I took a different route and showed how to find “yield” via the simulations I presented.
     I was trying to be very relevant – not “hijack” the problem posed. 
     Finally, I got interested in Six Sigma activities since I realized that the idea of a work function may be useful to these important problems facing modern corporations. Regards.
     Laxman 

    0
    #102970

    V. Laxmanan
    Member

    Dear RedFss:
    Let’s me stay focussed here.  This is what you had in your original post.
    If we assume the following opportunities, can we say still that no one among them is still near Six Sigma? Please correct my DPMO figures if you have them near to correct ones. They are just a wild guess.
    I have highlighted here what you assumed (a great start) and what you called a wild guess.  I got curious about these assumptions and the wild guess. 
    So, I did the simulation to arrive at the “yield”.  I also had to make some guesses along the way.  I had to define a “control limit”, like 3 defects per vehicle as the acceptable level of defects that Toyota’s customers have come to expect. Then I calculated Process Sigma using the yield as determined from the simulations.  The highest value I got was 3.55, not that far off from what you got (3.85 is the highest).
    But, there are still a lot of unknowns. If we could get the full data from the J. D. Power Surveys, I am sure it is lying around in somebody’s computer, we could actually test the whole thing – learn about a corporation from defects in its products, as revealed at the “macrolevel” . I hope this helps. Regards.
    Laxman 

    0
    #102971

    V. Laxmanan
    Member

    Dear RedFss:
    I mistyped 3.85 instead of 3.81.
    I also wanted to mention that y = hx + c has nothing to do with the simulation I did. You can do it yourself. Just distribute 207 defects among 100 vehicles and prepare a graph. 
    Actually, and on the contrary, the simulation shows that the simple law y = hx + c relates the number of defects y to the number of vehicles, as more and more vehicles are being observed. 
    Ultimately, this is also telling us something about the company – not directly, but indiretly. It all depends on how you would like to asses the performance of a company. I would really like to pursue this matter, if you wish, and the question that you have posed is a very important one. Regards.
    Laxman    
     

    0
    #102980

    Manikanda Mariappan S
    Participant

    Hi REDFSS,
    Without knowing the number of oppurtunities per unit for each type of car it is not logical to calculate six sigma capability using the reliability data.
    Ok, let’s say that you want to assume the number of oppurtunities per car and create a level platform to compare the sigma scores of each type. You are doing this just to get a rough idea of their processes.
    Why do you want to assume 200 for 1 type (toyota or honda) and 150 for another type (porsche) ? I feel that you should assume the same number for all types. In reality they may be different for different types of cars. But while we make assumptions, if we bias them, then we will not get a true picture.
    Say
    Number of Defects = D
    Number of cars (units) = U
    Number of Oppurtunities per Unit = Op
    Then, DPMO = (D/(U * Op))*1000000
    = (D*10^6/U)*(1/Op)
    In this expression D and U are variables. They differ for each type of car, Toyota, Honda or Porsche, etc.
    (1/Op) is a constant as we have decided to assume the same number of oppurtunites per unit for each type of car. So we can compare the value of (D*10^6/U) for each type of car to get an idea of their six sigma capabilities.
    Infact DPMO is by itself a very good measure of the capability of a process. Here (D*10^6/U) is equal to DPMO, but with Op=1.
    So if you happen to get the value of Op then you can get the real values of six sigma capabilities for the different car types.
    Thanks,
    SMM.

    0
    #102983

    Tater Salad
    Member

    REDFSS,
    Maybe you don´t understand a satisfactory answer when you see it.
    First point r/100 is not a reliability number.
    The data reflects escapes and doesn´t have anything to do with process capability. Customer data is a reflection of the efficiency of the defect detection and rework process. Actually the failure of it.

    0
    #102991

    Reigle Stewart
    Participant

    From benchmarking we know the average is about 4
    sigma. The automobile data is close. Measure with a
    micrometer, mark it with chaulk and then cut it with a
    hand-ax. Your data reveals a sigma that is below
    average. We know Toyota is better than average. So the
    defect opportunity count is much too low. I heard from a
    Ford executive there are about 2,000 opportunities in a
    car.

    0
    #102994

    Reigle Stewart
    Participant

    If we assume 200 defects per 100 cars, no matter how
    you cut it, that’s 2 defects per car. This is AFTER the
    influence of factory inspection and test. Assuming a
    combined containment efficiency of 85% (which is
    typical), we would see about 13 defects per car BEFORE
    test and inspection. Further assuming 2,000 CTQ’s per
    car, this gives us .0067 defects per opportunity (i.e., per
    CTQ), or a yield of about 99.33%. Since this yield is “long
    term” by nature, we can estimate the corresponding
    “sigma value.” This would be approximated as 2.48 +
    1.50 = 3.98, or about 4.00 sigma. If we consider 100
    defects per 100 cars, the resulting Sigma value is about
    4.2 sigma. On the other end of the spectrum, if we
    assume 300 defects per 100 cars, the capability would be
    about 3.8 sigma. This likely makes sense given the
    “average sigma” per part is about 4 sigma. From a “worst-
    case estimate” point of view, if we assume 300 defects
    per 100 cars, 95% containment efficiency, and only 500
    CTQs per car, we see a capability of about 2.67 sigma per
    CTQ. Best-case might be 50 defects per 100 cars with a
    containment efficiency of 50% and 3,000 CTQs, therein
    realizing a capability of about 4.9 sigma. So, we have an
    “extreme range” of 2.7 to 4.9 sigma with a “most probable”
    scenario of about 4 sigma. Remember, this is a
    “benchmarking” technique and it makes a lot of
    RATIONAL assumptions. It is merely a first-order
    approximation of the process capability. To gain more
    precision, one must dig to the next level of detail. But at
    the high level we are cruising, this approximation is “good
    enough.” We don’t need 7 digits of precision to make a 1
    digit decision. This is what I meant by the phrase
    “measure it with a micrometer, mark it with caulk and then
    cut it with a hand-ax. Reigle Stewart

    0
    #102999

    Frequent reader, infrequent po
    Participant

    Reigle,
     
    As a comment not meaning to inflame, just to observe, the last few days of response on your part to technical questions asked by forum members have been outstanding.   Please continue to provide your Six Sigma practitioner’s insights.   When you stay away from swapping personal punches with other members and answer and debate our questions on their technical merit – there are no better contributors.   Thanks for your help.  I (and others, I’m certain) appreciate it.   
     Frequent reader, infrequent poster.

    0
    #103011

    Mikel
    Member

    Dear Red,
    I didn’t miss any of the questions. I have watched with amusement while people who know nothing of the situation try to sound intelligent.
    Go visit Georgetown and then go see the Denso guys down in Marysville and compare to what you find in Detroit. You will realize the folly of your question. There is no comparison between Toyota and the big three. The battle is over, the big three are just taking a while to fall down completely.
    I am sorry you are not satisfied with the correct answer, I will do my best to ask you what answer you are looking for before responding again.

    0
    #103044

    PB
    Participant

    Stan,
    You are making a good point when you say the battle is over. Six Sigma application may not help solve ESCAPING QUALITY ISSUES. I strongly feel that car makers like Toyota and Honda (hopefully the big 3 will have learned this) have understood the power of long term reliabilty which equals long term satisfaction. One may have x defects per 100 cars but that may be cosmetic defects, or minor at best. However, I am sure that once they found that defect they will analyze their production line to find and eliminate that defect.
    PB

    0
    #103048

    Bev
    Participant

    REDFSS:  I did work at either Honda or Toyota (I know which one, I’m just not gonna say right now) in the nineties. 
    Bottom line to answer your fundamental question:  there is no way to determine internal sigma rates from external data.
    DPMO is relatively meaningless.  There are thousands of opportunities for failure on a vehicle (maybe tens of thousands).  There is some value in the manufacturer knowing its’ dpmo as new vehicles are introduced and quality improvements are made so upper management can make rational decisions when comparing defect rates for more and less complicated vehicles.  But the Customer only cares about how many defects they get on their car and that is the bottom line in the automotive industry and is the metric that counts.  So dph (defects per hundred cars) is used.
    MORE DETAILS
    One of my primary functions was to analyze the JD Power Initial Quality Survey results (published in May each year from new model year vehicles REGISTERED in the fall (I believe it was October & November). I then initiated corrective actions within the factory for those top items that  could be influenced by the factory.  The survey asked 90+ specific questions concerning defects that the customer might have encountered in the time since they bought hte vehicle and wehn they received the survey whcih as approximately 90 days after the last day of the registration period used to sample from.
    I am currently in another industry and so I am not familiar with the “relibility” survey, but it sounds suspiciously like the initial quality survey as the results were always published as defects per hundred cars.  The total number of ‘cars’ that responded were never published publicly, but were available to the automotie suppliers for a fee.  Suffice it to say the defects per hundred number IS the average defect rate (in pph) for that supplier.  Typically the defects are distributed among the vehicles with a skewed distribution…many vehicles have no defects, some have a lot.  (I know the best fit distribution for the vehicle I worked on, but obviously this will chagne based on each vehicle type and it’s maturity)
    As someone else pointed out very quickly, it is in fact impossible to determine the internal first pass yield numbers or internal sigma levels form external data…there is no universal mechanical (actual) correlation.  what is seen externally is a function of the internal defect rate and the effectiveness of the internal screens and rework process…and THAT is totally unique form model to model, factory to factory and year to year and automotive company to company. 
    This survey was my life for the time I was at the japanese auto maker…I always knew what my first pass yields were, my inspection escape rate, my rework effectiveness and my warranty claim rates – I could predict the JD Power score for my vehicle within 3 pph or better each year.  (I predicted teh score on the 4th model year wihtin 0.3 pph; it was less than 60 down from a high between 110-120 pph at on the first car of the major model I worked on.) 
    Other attempts to determine the internal sigma rates are just theories with no way to validate them unless you are on the inside and that is your job…the amount of data that must be analyzed is staggering.  YES, I know what my sigma rate was for my vehicle.  But I am prohibited from telling anyone.
    Obviously, the internal rate is greater than the external rate…so th sigma level is lower.

    0
    #103051

    Reigle Stewart
    Participant

    Bev: I would respectfully disagree with your position
    concerning the J.D. Powers data. Such disagreement is
    based on two points. First, it was Mr. Bill Smith at
    Motorola in the early 80’s that discovered an empirical
    correlation between the field failure rate of electronic
    equipment and the total defects produced during
    manufacture. Mr. Smith further postulated a correlation
    between the latent defect content of a product (flaws that
    require activation energy to made observable), total
    defects per unit (TDPU) and the field failure rate.
    Following this, Mr. Smith connected with Dr. Harry to
    further examine this association in the mid 80’s, but from a
    statistical and mathematical point of view. Subsequent
    emperical and simulation studies conducted by both
    gentlemen statistically ascertained that the “infant
    mortality period” and “useful life period” of the classic
    bathtub reliability curve are explained by three things;
    namely, existing design bandwidth, process operating
    bandwidth, and feature complexity. The interaction of
    these three factors governs the probability of observing
    an in-line defect (noting that defects are most often
    Poisson distributed). In turn, it was found that the latent
    defect content is directly proportional to this probability.
    From this, the two researchers concluded the
    mathematical association and subsequently published
    this information internally. Of course, the problem can be
    worked “in reverse.” The field performance data can be
    used to “back compute” the latent defect content which, in
    turn, is used to back compute the TDPU. Once the TDPU
    is estimated, the pre- and post-inspection performance
    yield of a “typical” CTQ can be approximated. Do
    recognize that such an approximation is just that, an
    approximation. In this context, the resulting “sigma” is
    merely a benchmark (not a precise measurement like
    would be ascertained from direct measurement of the
    related CTQs). Secondly, while deploying Six Sigma at
    Ford Motor Company, Dr. Harry and Mr. Phong Vu
    applied this reasoning and equations to the internal
    automobile quality data gathered at Ford and from the
    related J. D. Power’s reports. Using this data, they were
    able to successfully benchmark (and forecast) the
    inherent production capability of several vehicles.

    0
    #103056

    V. Laxmanan
    Member

    Dear Reigle Stewart:
    Thanks for sharing this insight.
    The matter is challenging and difficult but I think we have to continue to develop a method of correlating what is going within a large organization, like Ford as an example, with how it products are rated and/or perform in the real world. Unfortunately, good quantitative data to permit a rigorous correlation is lacking – but that should stop us. Regards.
    Laxman 

    0
    #103059

    howe
    Participant

    Great!! This is interesting! Two folks, you and Reigle, who don’t have any experience in actual implementation of Six Sigma as in NO SLOGANS but practical work will be able to solve the problems of “large organizations”. If you think these large organizations’ problems are technical, you must be naive.
    Here is an idea: Between you two, you can start a new company and call it Six Sigma GEN Infinity!

    0
    #103060

    Mikel
    Member

    Reigle,
    Here we go again. Bev has data and experience. You have opinion and Dr. Harry fables. You have no actual knowledge of Toyota or Honda – if you did you would understand the importance of Lean (you do not) and you would understand the folly of your post. Ford had a double digit sales decline last month and year over year declines. The quality and reliability data shows Honda and Toyota to be far superior and if you knew how they do business, you would know they do it with very little test and inspection as we know it.
    Respectfully, you don’t have a clue and all of the Dr. Harry stories in the world will not change that.

    0
    #103066

    V. Laxmanan
    Member

    Dear Mike:
    I have been in large organizations all my life and I have actually been involved in implemetation of R & D programs, aimed at producing cost savings throughout the corporations. Just so you know where I am coming from. Regards.
    Laxman 

    0
    #103115

    Dorf
    Participant

    Reigle – I respectfully disagree with your statement about successful prediction of vehicle process capability.  Putting it on a Powerpoint slide does not make it so.  Mr. Phong Vu had a very tough job at Ford, and he arguably did the best he could with the pressure he was under, but the program fell into the trap of having to justify the cost by declaring success on a schedule.  Once you start counting the number of BB’s as a measure of success, you have fallen into the consultant trap – the bigger your program, the bigger your success, the faster your SS program ramps up, the faster you save billions.
    There are very capable engineers at Ford – as good and better than Toyota and Honda in many cases (that statement based on participation in many competitive teardowns & supplier interactions) –  and very good quality and reliability processes – the Management resolve and responsibility for results, action & progress is absent at critical levels.  Setting up rigorous Engineering Gateways does no good if Management blows right through them. Watching fast-track Management promotions with declarations of success, while your relatives have the transmissions in their Taurus’s and Windstars fail before the first year of ownership is over – year after year after year -well, that tells the story better than any Powerpoint presentation of Capability prediction.

    0
    #103120

    V. Laxmanan
    Member

    Dear Dorf:
    Please allow me to put what you have said slightly differently.
    When you focus your attention on the Internal measures of process capabilities, and Process Sigma values, and Six Sigma Implementation Programs, you are falling into what you have called the consultant trap.  The more the BB the bigger the costs savings and the bigger profits etc. I agree.  
    But, when you start looking at the External measures of your product is performing, which is what the J. D. Power Surveys tell us, then you get a very different picture of the corporation. It might not agree with Six Sigma practitioners are taught about Process capability, 1.5 sigma shift, etc. But, that is the stark reality. 
    If Toyota is beating your pants off in these surveys year after year, you will not be able to sell your vehicles because customer “perception” of quality is very important.
    And so it is that I tried to do what I did in the last few days, but looks I am just a transparent belt (like someone put it) not a seasoned BB, or MBB, etc.
    Getting great numbers for the Internal Six Sigma is like a student getting A’s and a couple of B’s may be in all the high school courses.  But when the student takes the SAT he or she doesn’t do so well – and, unfortunately, performs more like a C grade student. The SAT here is the external measure of performance and one that many college use, along with the internal measure (the high school GPA).
    In the case of a company, I don’t think regular customers who use the  finished product (like those who respond to the J. D. Power Surveys) are not paying attention to any Internal Process Sigma values. Regards.
    Laxman

    0
    #103124

    newbie
    Participant

    Great post and exactly the kind of information I have been looking for.  What is most interesting is that the 2.7 defects per unit in my world would get me fired. Interesting that Toyota gets to state they are world class.  I would be interested in what paper was the source for this information. I work in an environment where irregardless of the number of opportunities 1 error is counted against the end product, as such 4 defects per 100 items would equal 96% quality. The logic is that 4% of our customers recieve an end product not meeting each and every expectation.   No consideration is given to the fact that there are 30 opportunities for defects to occur on each item. 

    0
    #103126

    Reigle Stewart
    Participant

    V. Laxmanan: Your position has lots of emotional appeal
    but might be lacking in the operations management
    arena. I would agree that external measures are the end
    game. After all, it is the customer’s perspective.
    Nontheless, we must also have internal measures of
    performance, especially those measure that correlate to
    the external measures. Merely looking at external
    measures is like trying to steer your boat by looking at the
    wake. Of course, internal measures must be statistically
    correlated to the external measures. When such
    correlation exits, we then have “radar” to “see” where we
    are going. By identifying and verifying the key internal
    measures, we better understand what “knobs” must be
    turned to make things better and keep the ship on course.
    Respectfully, Reigle Stewart

    0
    #103129

    Peppe
    Participant

    Dear Reigle,
    you are correct, but mainly about early life failure. Take care that Toyota, in Europe, was the first car maker that gived 5 years of total waarnty, also on cars with price less than 12.000 $. You are too expert to understand what this means in terms of factory failure, early life failure and long term reliability and Toyota was also the first car maker to give 3 years of full warranty, in europe,  5 years ago.
    All people in this forum have the right experience to understand what means to move from 3 years to 5 years of warranty (+60%) within 4 years having the same price of 2 years ago. I don’t know how many other car makers had did it.
    Rgs,
    Peppe
     

    0
    #103132

    V. Laxmanan
    Member

    Dear Reigle Stewart:
    I agree. We need both internal and external measures.
    A company with no good internal measures will not be able to deliver anything of value to the customer. That is where Toyota may have succeeded.  By rigorously instituting such internal measures, they have produced good external measures – the end game is what you now see in the J. D. Power surveys.
    Now, the question that still remains: How do we correlate the internal with the external?  I cannot speak much for the internal measures at Toyota, Ford, or GE, or even GM. But, I have certainly studied the external data, such as the J. D. Power Surveys.  This leaves much to be desired, as I have tried to show with my “simulations” of how defects might evolve to the DPV (defects per vehicle) levels given in the J. D. Power surveys, when we consider just 100 vehicles. 
    Several thousand responses were used to develop the reported DPV values. We don’t get a “fair” evaluation of a company from the y/x ratios since the governing law is clearly y = hx + c with x being vehicles tested and y the number of defects. We must know both x and y, not just the ratio y/x.  The J. D. Power surveys, in my humble opinion, should be reporting both the x and y values, not just the ratio, or the per 100 vehicles numbers as is now done.
    However, please, this is not meant to be a criticism of the J. D. Power surveys, or the now widespread use of simple y/x ratios as measures of performance or efficiency.  I don’t think anyone has so far paid much attention to the fact that if y = hx + c, then the ratio y/x = h + (c/x) which means the ratio y/x is NOT the same as the rate of change h, unless the intercept c is zero exactly.  Now, the intercept has been renamed the work function – and for good reasons. That’s all. 
    We must pay attention to the nonzero intercept, or the work function, why it is what it is and how it is changing or else we draw misleading and even erroenous conclusions with the y/x ratios.  I certainly hope we can ALL agree on that.
    It is much too important, in my humble opinion, for the longer term success of Six Sigma Implementation Programs.  Again to use the student analogy, if a student gets a failing grade on the SAT (or less then sterling performance that was expected) but has very good grades in high school, one must wonder why.  Colleges are forgiving and accept the internal grades along with the external grades. But customer may not be, since they have no clue about what is going on internally.
    At present why corporations may be “failing” because the external grade is simple y/x ratio, which may not be the “true” measure of performance. As those who believe in Six Sigma methodologies, we  have to be concerned about this as well. Regards.
    Laxman
     

    0
    #103134

    V. Laxmanan
    Member

    Dear Newbie:
    I am not clear if you wanted me to respond. If so, the information that I have analyzed was obtained from the J. D. Power website. If you had something else in mind, please accept my apologies in advance. Regards.
    . Laxman

    0
    #103157

    Reigle Stewart
    Participant

    Peppe: You are right on target. As the mean of each CTQ
    approaches its target value (nominal specification) and
    the variance is reduced, the probability of defect is
    reduced. As the probability of defect is reduced for each
    CTQ, the rolled-yield is increased. As the rolled yield
    increases, the the total defects per unit is reduced which,
    in turn, leads to a reduction in the latent defect content of
    the product. Given a constant combined test and
    inspection efficiency, the escape rate is reduced. As the
    escape rate is reduced, field failures is reduced. As the
    field failure rate is reduced, the warrenty period can be
    extended. As further improvement in process capability is
    realized, there is less need for test and inspection. As
    rolled-yield increases, cycle-time goes down. As cycle
    time goes down, work-in-process is reduced. Thus, we
    have the “micro economics” of Six Sigma. Respectfully,
    Reigle Stewart

    0
    #103180

    Cravens
    Participant

    Are you suggesting that Toyota has used Six Sigma as the method to achieve the results referred to by Joseph?

    0
    #103183

    Peppe
    Participant

    Dear Reigle,
    your explanation is, of course, absolutely right. Just one point : many time, here in this forum, was stated that Toyota doesn’t use SS, but TPS, but, at least, are we talking about same things (kaizen, lean, SS, TQM, TPS) ? 
    An information I whish to know, if available, from how many time Toyota and Big 3 are started with TPS and/or SS programme ?
    I think this is the answer about differencies.
    Rgs,
    Peppe

    0
    #103268

    Hemant Gham
    Participant

    I agree that there needs to be an attempt (if not now, maybe sometime in future) to correlate internal quality measures to external quality measures. The principle of cause and effect will always keep controlling the businesses even if we try to ignore this. There needs to be a relationship, a linkage established between external measures like customer satisfaction and internal measures that reflect the true performance and reliability of internal processes.
    I went through this interesting thread and it made me post some remarkable facts of Toyota. In Toyota the management systems make a real difference along with thinking and philosophy.  “Continuous Improvement” and “ Respect for People” are two cornerstone values for Toyota’s success. Toyota believes in continuous learning and has created an environment within the company that not only makes people accept change but willingly embrace it. How many companies have actually done that and been successful? In 1990s all of the Big 3s realized the rising of Toyota and decided to beat it by studying and creating their own versions of Toyota’s systems. They benchmarked the company on its production system, product development system, supplier relationship management system and others.
    Change management was added to Six Sigma in latter stages. Toyota should be having that quite sometime back I suppose. But what makes difference in Toyota and what I strongly believe in is the 14 principles that are ingrained in the processes executed in the company. While we can find what they are in many publications, the critical differentiating factors are “Long-term thinking”, “One-piece flow”, “Organizational Learning”. One-piece flow can surface problems that demand fast solutions. People are given the sense of urgency required to face business problems. In short, the management builds people, not just cars. As far learning is concerned right combination of philosophy, process, people and problem solving can create a learning enterprise.
    In 1990s Toyota came to limelight by engineering and making vehicles that led incredible consistency in the process. It designed vehicles faster, with more reliability, yet at competitive cost. When some defects appeared Toyota miraculously fixed the problem and came back even more strongly.
    Some statistics:
    ·        Toyota is 3rd largest auto manufacturer in the world, behind GM and Ford.
    ·        Global sales = 6 million per year in 170 countries. (If this trend continues, it will pass GM in 2005).
    ·        Annual profit (march 2003) = $ 8.13 billion –larger than combined GM, Chrysler, and Ford. (Do profits declared by these three have Six Sigma financial benefits?)
    ·        Increase in Toyota’s shares = 24 % over 2002.
    ·        Market capitalization = $105 billion (2003).
    ·        Vehicles sold in North America = 1.8 million in 2002 (Redfss can work on these numbers). As far as DPO is considered, it is too difficult I think to obtain the number, but we can use the quality checklist for every finished vehicle. For example, 170-point quality check for every finished Lexus RX 330 can be considered or average if we have data on other models.
    ·        79% fewer Toyota vehicles were recalled in the US than Ford and 92% fewer than Chrysler in the U.S. (2003).
    Toyota has the fastest product development process in the world. Is DFSS integral part of TPS?
    Toyota employs quality improvement methods such as just-in-time, kaizen, one-piece flow, jidoka (Built-in quality), and heijunka (Level out the workload). It has a true balanced lean flow of work. The three M’s (Muda- Non-value-added activities; Muri – Overburdening people and machines; Mura – Unevenness) form the concept behind heijunka. And many companies that have Six Sigma rolled-out in every department do not have anything in parallel to them. It took U.S. auto companies years to understand how to apply “Andon” system – “fixed-position line stop system”.
    J.D. Power’s survey of initial quality show that the gap between actual Japanese auto companies and U.S. competitors has reduced, but long-term data shows that there is a big difference, which is hidden. Initial quality (during 1st three months of ownership) shows little difference, but three years out, the gap grows.
    The data, which REDFSS posted, was I think for vehicles three years old (sold somewhere in 2001/02??) During this 1st three years of ownership.
    Toyota = 207 def / 100
    American Honda = 210 def / 100
    Porsche = 237 def / 100
    What I have heard is that Six Sigma is there in Toyota. Still I could not get the exact need of it. There are Black Belts but was told that they keep things simple and do not use complex statistical tools. The 4 key tools used to attack problems are:
    1.      Go and See.
    2.      Analyze.
    3.      Use one-piece flow and “Andon” to highlight problems.
    4.      Ask “Why?” 5 times (integral part of kaizen) to get root cause as well as countermeasures to solve it.
    Also, who can forget Toyota’s supplier relationship management, and 1997 incident when one of the biggest and closest suppliers failed to deliver critical part used – at that time 32,500 per day. 200 suppliers had self-organized to get the production of the parts started in 2 days, rigging temporary lines to make them, and kept business unaffected. This power of supply chain is far more than Information Technology.
    Have a look at the Toyota’s 7 steps practical problem-solving process. Tools and techniques are used, but more emphasis is on thinking through problems and solutions. At Toyota, it is said that problem solving is 20% tools and 80% thinking. I am sure as a Six Sigma practitioner what I have experienced is that many companies get caught up in using all the great and new sophisticated analysis tools with problem-solving is 80% tools and 20% thinking.
    Finally, at the same time Six Sigma was spreading, many companies were applying lean tools and having some success with that. Six Sigma can surely improve value-added process-like finding the quality problems and fix it, but lean focused on the whole value stream and creating flow among the value-adding operations. Six Sigma fixes individual processes not connections among them. New hybrid – Lean Six Sigma, well I am not aware it is working. Toyota has aggressive internal targets, which I guess are taken under Six Sigma projects banner.
    I still wonder why TPS should let in Six Sigma. Many views are expected on this.
    Hemant
     

    0
    #103269

    Anonymous
    Guest

    Hemant,
    I enjoyed reading your post … something you might like to reflect upon is the ‘complementary nature’ of ‘long term’ and ‘short term.’
    Sometimes, very short term activities have the effect of promoting  ‘long term’ views … such is the wisdom of oriental philosophy.
    Andy

    0
    #103271

    Mikel
    Member

    Hermant,
    Good post, good job of getting facts – until the end. It is a rumour, not a fact, that Toyota is doing Six Sigma.
    Also a minor point – change management has always been part of Six Sigma.

    0
    #103280

    PB
    Participant

    Hemant,
    Good post. I believe in simplification and not overcomplicating and this post is right up my alley.
    PB

    0
    #103297

    art gonzales
    Participant

    Hi REDFFS,
    I tried to go back on the emails (up to  July 6) and see no clarification on J.D. Power survey procedure. For the benefit of readers like me without good reliability  backgrounds, is the J.D. Power survey meant to be a reliability measure? Is  the “long-term quality” ( ? 3 years) survey meant to measure customer satisfaction over this period, really a valid measure of reliability? If reliability is  the probability of performing a function over some time under certain conditions, is the JD Power survey figures a valid  measure of reliability ? Or is the JDP survey figures meant to measure ” percieved quality on the long run” ? Are reliability and “perceived quality on the long run” the same?
    Thank you,
    Art G
     

    0
Viewing 56 posts - 1 through 56 (of 56 total)

The forum ‘General’ is closed to new topics and replies.