iSixSigma

How to Determine If the Process Is in Control?

Six Sigma – iSixSigma Forums Old Forums General How to Determine If the Process Is in Control?

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #35639

    Arora
    Participant

    I want to know how the process could be assumed to be in control if it meets the set of criteria(like no points in chart is above the 3 control limits, six successive points that increase or decrease,four out of five points that are on the same side of the centerline, etc)
    Say for eg, I have set a target for CSAT as 4.5. I have data of about 20 subgroups with the subgroup size of 5.I calculate UCL,LCL,RBAR and draw the control chart.All the points are within the control limits.But, none of these points are ouching 4.5.
    How then can we say this process is in control?

    0
    #100702

    Gabriel
    Participant

    To say “it is in control” you compare the process against itself. If its behaviour is consistent over the time then it is in control. You don’t even need specifications to see if it is in control or not.
    When you compare the process output against a specification, then you are talking about process capability or process performance.
    Even when a good capability is needed, tyipcally stability (another way to say “in control”) is needed first. If the process is stable, you can compare its performance against the requiered performance and take corrective actions if needed. But if it is not stable, you can hardly even compare the process against something, because a thing such as “the process” does not even exist from a statistical point of view, as its behaviour is changing over the time so you don’t have one distribution to medel the process.
    For example, if the process is stable but not capable you can predict that you will have let’s say 20% scrap. This can be not acceptable but you know what you will get, where you are and where you need to steer to. If the process is not stable, then you don’t know what you will get, where you are, and where to steer to, except that you need to stabilize the process first.
    Just my opinion.

    0
    #100707

    Arora
    Participant

    Excellent!
    Thank you Gabriel. I understand that the difference is “Stability” and “Capability”.
    When you have a bunch of data , whats the best way to go?.
    1.Find existing stability in the process with the control chart.
    2.Find the process capability
    3.after process improvement, again find te process stability and capability?
    If the process stabiliyt is poor due to special causes, how will you proceed in the first place?

    0
    #100716

    Pallab
    Participant

    Kris,
           As Gabriel rightly explained, there are two behavioral parameter of a process.
    1. Stability(caused by process variation or consistency issues and measured by control chart)
    2. Capability(response location with respect to the specification limits, caused by process centering issues or accuracy issues)
    Now the biggest question is which one to get priority over the other ?! The answer is “stability”. First if the process is not stable, the corrective measure will have to be taken through Root cause analysis or Six sigma data analysis. This is the time to handle special causes. Rigorous statistical data analysis comes with solution which may 1. shift mean and 2. reduce the variation. As this stage, the process improvement goal is to reduce considerably standard deviation of the process with small or zero improvements in process sigma(capability), as the process is not delivering at all within spec limits. Here, you may use DoE or Taguchi’s orthogonal parameter design to focus on outliers and its root causes.
    2. Stege 2 is after successfully achieving stability, you have a process at hand which listens to your instruction, it only varies with random factors and has become consistent. This is the best time to push the process towards spec limits and make the process capable. Here you may look for technology change, e.g. One web portal was failing to meet customer performance CTQs, you had already undergone exercise to make the page load time consistent by reducing number of database calls, using less overhead on server more on client and fine tuning the querries. Though, after having these improvement solutions, mean response time is 10 sec and very consistent around 10 secs. Where your spec limit is <= 5 sec. One of the innovative solution using TRIZ (theory of inventive problem solving) is, shift the architecture of the portal from ASP to JSP with EJB(from microsoft to J2EE using enterprise Java beans). As a result, the mean response time came down to 4 sec. Now you can start drawing control chart to check the "capable" process is stable or not.
    Note : Very ofthen this happen that, a mean shift brings in  “free”(unwanted) inconsistency in the process. So, we need to be aware.
    Hope this helps.
    Regards.
    Pallab, TCS MBB

    0
    #100720

    Arora
    Participant

    Thanks Pallab!
    Great post!It did help me.Yet another question in your example.You had mentioned that your mean response time was about 10 secs consistently against the specs.But, what about a scenario when your mean response time is fluctuating from 1-12 with randomness.I understand , you would first aim to to rest soemwhere to start finding the process capability.But, how will you go about in making this process stable first.There are so many special causes points in control chart.
     

    0
    #100741

    Gabriel
    Participant

    If it “fluctuating from 1-12 with randomness”, and only “with randomness” then it is stable. The fact that what your process does not give you what you want does not mean that it is not stable. Just that it is not the process you need. Change the process.
    Note 1: “Change the process” can be as dramatic as “trash the machine and buy a brand new one” or as simple as “move that knob one poit to the right” or “use your right hand instead of the left one”.
    Not 2: Note that, by introducing a change in the process, you are being a special cause of variation as you are changing the behaviour of the process. Being “improvement” a sub-type of “change”, stabilty and improvement are contradictory. Stability is a good thing when you want to get and keep the best from what you have now, not when you want to improve it.

    0
    #100742

    Gabriel
    Participant

    Be careful with your “bunch of data”. Remember that a control chart is a chart of some statistic (vertical axis) against time (horizontal axis). The data in the control chart must be time-ordered, and each point in the chart represents one instant, so if eahc point represents a subgroup then all the individual of the subgroup should be done is the shortest time possible.
    That means that if you take a bag of balls, measure a number of them and record the data, you can make several thing with that data but NOT a control chart.
    Otherwise, your 1,2,3 steps process looks pretty well.
    “If the process stabiliyt is poor due to special causes, how will you proceed in the first place?”
    The easy answer is “plot acontrol chart and investigate each occurrence of a special cause that shows on the chart as a signal, then take actions to eliminate one by one the special causes of variation  unti you get a stable process”.
    This is what would apply to the typical cases. However, a more correct answer would be: Understand the special causes and their effects on the process (a control chart can help). Then decide what you want to do with them.
    Not allways all special causes must be eliminated. Some times one would want to but it is impossible with the current state of the arts. Some times one would want to eliminate the special causes and know how to do it but it is too expensive or too complicated and there are other priorities. Some times one knows that the variation due to a given special cause is such that it will never put the quality in risk, so one does not need to remove it. Sometimes one can live with an unstable process if one knows why it is unstable and knows the extent of the effect of the unstability.

    0
Viewing 7 posts - 1 through 7 (of 7 total)

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