iSixSigma

Targeted Mean or DPPM to be Project Indicator?

Six Sigma – iSixSigma Forums Old Forums General Targeted Mean or DPPM to be Project Indicator?

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #32328

    Cactus
    Participant

    Dear all,
    I got a puzzle when I try to determine the target for a project which aim to reduce the Cycle Time of a transactional process.
    I might have 2 ways  to define the project goal.
    Way I:  Reduce the Cycle Time Mean from Baseline to a target.
    Way II: Set a acceptable cycle time (Spec.), meanwhile calculate baseline DPMO, the project goal is to reduce the DPMO to a new level.
    Can anyone advise me which definition is better?
     
    Thanks.
    Cactus
    May 23
     

    0
    #86241

    Mike Carnell
    Participant

    Cactus,
    I’m not sure what the difference is. It seems like a waste of time to do DPMO. If you have a team (which you should) it would seem less complicated to talk to them in terms of time rather than explain the DPMO idea.
    You are going to have to shift the mean regadless of how you measure it. You need to check for homoscedasticity so when you shift the mean you don’t inflate the variance. If it is a machine the mean shift tends to be a knob variable. If it is people it will do all kinds of stange things so it is more complicated.
    Just for simplicity I would stay in time measurements and watch the mean and std. deviation.
    There is another measurement we use called entitlement. Basically “the best you can do with what you have.” You time just the value add stuff that is being done. It help to know that number so you don’t set your target below that number. Dropping entitlement by definition takes a process change.
    Are you running to a takt time target? I guess the real question is where is the target coming from?
    Good luck

    0
    #86244

    Tierradentro
    Participant

    Cactus,
                I have completed various cycle time projects. I set a target and then calculated my sigma value against a capability analysis. If I said the target was ten  days I wouild then collect the data and figure out what is the percentage yield against ten days. Beware that the data could be lognormal so do a box cox and then capability analysis with usl of 10 days.
    In my organisation we always used to measure cycle time with the mean but I have now educated people into the understanding that the data is skewed and we all now measure the median.
    So percentage yield of target, = Sigma Value Improvement
    Hope this helps,
    john.
     

    0
    #86335

    Kannan
    Participant

    Hi Cactus,
    This is a question I often used to ask myself in my early days as a black belt. It almost feels when you finish the project charter you have completed measure phase. But the truth is if you are convinced there is a problem you have seen at some data. So you need to be able to make a Problem Goal Statement that is a combination of Specific, Measurable, Attainable, Relevant and Time bound (SMART). Even if you don’t go to DPMO measurement – you atleast need to have a flavor of # of Customer Complaints due to delay in your processing, The Pain that is caused and any number associated with the pain. Measure Phase has Performance standards where you come up with a goal again and Analyze has Performance Objectives after you have arrived at process entitlement…so it is iterative and what you find after current process entitlement is your final Goal. Understand the relevance of SMART Problem / Goal, then do Performance Standards and finally Performance Objectives.
    Cheers

    0
    #86377

    Holly
    Participant

    Similar to John’s reply, I have found that cycle time data is typically skewed and thus use median vs. mean.  To add to that however, we all know that when looking at continuous data, there are two measurements, typically mean and standard deviation where the standard deviation is an indication of stability.  If the data is skewed, what do you use?  What I learned and have found useful is to use the median and “span”, where span is the difference between the 5th and 95th percentile, or sometimes the 10th and 90th, depending on the number of data points.  The reason to use span is that in most cycle time cases, the customer is more apt to be affected by variation than the “average”.  Bottomline is 1) when you have continuous data, try to use the continuous data and stick with cycle time, and 2) find out if your customer is really affected by the average, or the stability.  If stability is the issue, then make your goal to reduce the span or another indication of stability.   

    0
    #86422

    April Brown
    Participant

    This is quite a familiar topic to me as I am just finnishing my first cycle time project.  A targeted ( specified) cycle time would work best for a transactional process. Question- Do you have any historical data to support for specified target? Let me know what you come up with

    0
Viewing 6 posts - 1 through 6 (of 6 total)

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