iSixSigma

SPC and the Real World

Six Sigma – iSixSigma Forums Old Forums General SPC and the Real World

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #30122

    James A
    Participant

    For some years I’ve been pondering the age-old issues around SPC charts – you know the sorts of things – ‘creative’ measurements, filling in the charts at the end of the day, creating charts to keep ‘management’ happy etc.
    The question is this – if there is a suspicion that ‘creative practices’ are at work on a chart, is there a tool within MiniTab that can help to identify (statistically) the probability that the data is not real?
    Has anyone else come across similar issues – and if so – how were they solved?  Training I know is not the answer – and use of roving inspectors is not VA and not effective.  Standing over operators would not a happy workforce make, and sacking everyone would slow production and introduce a culture of fear.
    I’m sure we don’t have this problem here – but I have come across it in the past – and could do with a few pointers in case I come across it again!
    Billybob (and anyone else out there) – you got any non-possum/non-raccoon-involving ideas?  Aversion therapy using cattle-prods seems viable – but there must be a better alternative.
    James A

    0
    #78137

    Marc Richardson
    Participant

    Hi James,
    I went to my first SPC seminar in 1976 and have been using it ever since. I have run across this issue many times. First of all, let me say this: when I encounter what appears to be measurement discrepancies, I do not make any assumptions about the person’s motives, that is, I do not assume they are pencil whipping the data. If, in fact, they are, it is because the system either rewards them for doing so or punishes them in some way if they don’t. Either their workload is too great to allow them sufficient time or their supervisor will criticize them for unfavorable measurements, etc.
    That being said, the first step prior to implementing an SPC application is to certify the measurement system by assuring it is calibrated and conducting bias, linearity, stability repeatability and reproducibility studies. This means using the personnel who will be performing the measurements, the actual measuring equipment, parts and measurements that will be measured during production for the Gage R&R studies. Further information on this is available in the Automotive Industry Action Groups “Measurement Systems Analysis” manual (http://www.AIAG.org).
    Once the measurement system has be certified as being capable of providing measurements which do not significantly add to the process variation, the charts themselves will tell you what you need to know. An example of this is a chart involving the measurement of the length of a part. We noticed that some of the measurements seemed to be out of line so we sorted (using our nifty SPC software) the data by operator and redisplayed the control chart. Sure enough, one particular operator was consistently getting longer measurements than any of the other operators. Further investigation revealed that the operator was measuring the part a slightly different point. We revised the work instructions to clarify the where the measurement should be taken and the problem was solved.
    Best Regards,
    Marc Richardson
    Sr. Quality Assurance Engineer

    0
    #78140

    James A
    Participant

    Hi Marc,
    Thanks for your response – but let’s assume we’ve done all the GR&Rs, we’ve done the work studies, we know the FTF times, and we know the capbility of the machines, and we are depressingly familiar with the good words from aiag, aiob, IATF, QS, TS and Uncle Tom Cobbly and all.
    Let’s also assume that there is no criticism for out of control points, that continuous process improvement is encouraged, and that the leadership promotes ‘Authority to Stop’ where needed.
    Like I asked – is there a tool within MiniTab that would assist in identifying – with a reasonable degree of certainty whether or not ‘pencil whipping’ as you put it, has been or is occurring?
    Thanks for your response – and sorry for the delay in replying – I’ve been off getting my hands dirty – honestly, who’d want to be an engineer (only joking).
    James A

    0
    #78142

    Mike Carnell
    Participant

    James,
    Anonymity is at risk when you use words like “sacked.” That kind of narrows things down some.
    The racoons had no particular position on this problem. Actually they are sleeping.
    The situation you describe is, as Marc said, a function of something else. If people are doing this they are trying to survive in a system that is not responding to a OOC, getting some crap over shutting the line down, etc. I have run into very few people on a factory floor who do not want to do the job not only right but well. When they behave in some fashion that isn’t consistent with this you really need to look at the environment.
    In general the basic rules for SPC will tell you when the data is behaving badly. First there should be occurances of OOC with no particular cause – so if it is never OOC then something is wrong. If they are using a set of parts that they know will pass, you will get a value that repeats, etc.
    The one I had the most difficult finding was a product that had to reach a certain weight to pass. If it didn’t pass they didn’t get a label for the box. The operators were so skilled they could tell as the scale settled on a number that it wasn’t going to make weight and they kept some clay in their pocket and knew how much to pull off and drop in the scale to get it to make weight. The only way I found it was by working in the operation. I was the only one reject product so one of the guys showed me how to get them out the door. We drove the behavior by watching amount of product packed out. We changed the metric to how much they processed regardless of packed or rejected the problem stopped because it didn’t matter any more.
    Behavior is always a function of something else.
    Good luck.

    0
    #78145

    Erik L
    Participant

    James,
    I’ve seen two different tools used to identify potential situations like you’re mentioning.  The first is using the SPC charts that you’re mentioning.  Enabling a greater sub-set of the Western Electric rules could help to potentially unearth pencil-whipping.  Of course, the more you turn on the more you increase your alpha-risk.  I guess it depends on how ready you are to chase potential phantoms through the process.  The second tool is looking at the distribution of the data coming off the process.  Stack-ups at particular values, or truncated distributions are typically signals that there’s some fudging going on.  A combination of the two would probably be the best.
    Regards,
    Erik

    0
    #78176

    James A
    Participant

    Firstly,
    Thanks to all who responded – like I said, I’m sure we don’t have these problems here – but one must be ever vigilant – if we can’t cope with truth in engineering, then a lot of people are going to be disappointed in their product!
    Skew and/or kurtosis seem to be worth having a look at some day on some ‘genuinely pencil-whipped’ examples.  Like most of you alluded – changing the environment seems to be a key in this as well – and as has been shown many times on this forum, that’s the hard part.
    And yes, Mike, I freely confess to using non-PC phraseology – but at the end of the day, it doesn’t matter whether you refer to it as being RIF’d, downsized, re-allocated out, transferred, moved sideways, or plain fired – it’s basically all the same thing – and we know it – no matter what country we’re in.
    Thanks again for the input – all of you.
    James A

    0
    #78177

    James A
    Participant

    I’ve just caught up with this comment in another thread from Anna O’C (the now free, former Annonymous) and I think it sums things up pretty well:-
    “Organizations have “homeostasis” just like organisms do.  So they try to return to their previous state.  That’s why change is hard unless the organization or system is deliberately desdigned and trained to accept change.  That’s why eternal vigilance is the price of real improvement / process control.”
    Maybe that approach is what we should all be aiming for in systems design and 6S projects – thanks for that Anna – and yes, eternal vigilance I totally agree with.
    James A

    0
    #78190

    Mike Carnell
    Participant

    James,
    Didn’t mean to infer that sacked was not PC. If you are shooting for some level of anonymity it narrows down the world to a few places.
    Somewhere I missed to quote from Anna. It was great. Thanks for bringing it up again.

    0
Viewing 8 posts - 1 through 8 (of 8 total)

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