iSixSigma

Control charts for passive processes

Six Sigma – iSixSigma Forums Old Forums General Control charts for passive processes

Viewing 69 posts - 1 through 69 (of 69 total)
  • Author
    Posts
  • #53174

    Marcy
    Participant

    I’m stumped.  We have a process which we have no control over.  We count the number of forms that come in every month, but we have no control whatsoever over how many come in (someone has to die first).  We are basically a big passive receptacle.  A coworker, a brand new BB, wants to use control charts to illustrate these numbers.  Since we have no control, how can a control chart assist us?  If it goes in or out of the UCL or LCL, there’s nothing we can do.  I think a simple time series makes more sense.    Any thoughts? 

    0
    #188663

    Helper
    Participant

    Listen to your black belt.  You must first bring your processes under control in order to improve quality and productivity.  1st step is two measure stability, in a statistical sense.
     
    Listen to your black belt….

    0
    #188664

    Tierradentro
    Participant

    The CC tells you if your process is in control (statistically speaking) and has little to do with the degree to which you currently control the process.  The choice of time series plot (CC vs Run) has nothing to do with the degree to which you think you control the process either, but rather the limitations of your data.
    Go back and charter this thing up and determine what you are trying to accomplish. CC comes later.  Good luck.

    0
    #188665

    Mikel
    Member

    WrongI’ve improved many processes that were not under control.It is not a precondition.

    0
    #188671

    Marcy
    Participant

    Hi, Marcy the original poster again.  I did not clarify the process well enough.  I am also a BB.  The process in question cannot be improved in any way.  It can only be counted and reported.  I work in the death benefits office for a retirement fund administrator (401k).  When one of our participants dies, a family member sends in a specific form, w/ a death certificate, and with that form we determine how to disburse the funds.  Some months we receive more forms that others, but we have no control over the deaths of our participants or the number of forms we will receive.  We’re just a passive receptacle for forms.  Therein lies the rub.  The new BB is pushing hard to show these numbers in control charts, but I say UCL and LCL are meaningless in this case.   I hope that makes it clearer.  I can make a control chart just as easily as I can make a run chart, but I think the format is incorrect.  Thanks.  
     
     

    0
    #188673

    ?
    Participant

    Thanks for asking this questions it nice to see how many people actually read the question, which from you answer isn’t many.
    If your other measures are displayed in CC format I can see a use for it on a consistancy front, also the rules may help but probably not other than getting an upward trend when it’s really cold or hot. I’m guessing you’ve already checked your input data fluctuations against customer base / weather/ holiday season etc to check for trends.
    My guess is that the BB it wanting to cover their back by giving themselvs a get out of “the inputs aren’t in control how can you expect me to get the outputs”. Which doesn’t mean anything but that would depend on the knowledge of the people you’re working for.
     

    0
    #188674

    Robert Butler
    Participant

      So what you are saying is that when and how people die is completely random and when and how people fill out the form is completely random and there are absolutely no issues with respect to filling out the form so that any and all errors/delays associated with filling out the form/filing the form are completely random and there are absolutely no issues with respect to disbursment of funds – all funds of any kind are equally likely to be disbursed and any problems with the disbursment of any kind of fund from any source is completely random.
      I suppose this could be the case but I’d want to be able to prove it. 
      If I were given the task of looking at your process I’d start with a simple time plot just to see if there is any temporal pattern to form submission (greater numbers as the weekend approaches, etc.).  I’d start a bean count of any kind of issue at any phase of the process to see if any patterns emerge (disbursment of certain types of funds taking longer, more prone to filing error, etc. issues concening filling out and filing of form – certain fields more likely to have errors that delay the process, filing under some circumstances (socio-economic, regional, etc.) which seem to result in greater delay time, etc.
      Armed with information such as this you will be in a position to characterize your process.  If your analysis shows nothing but random variation well, you will know that for a fact and you can press on with other issues.  If it shows something else then you may have identified things that, if changed, would lead to process improvement.

    0
    #188675

    MrMHead
    Participant

    Is there an upper limit to the number of forms you can successfully handle in a specific (required?) time period?
    If you detect that, can you adjust something downstream to accommodate?
    Is there a lower limit where there is too much idle time / not enough work which wastes time/money?(correct me if I’m wrong)
    Control limits are based on process variability and specification limits.
    Where did the specification limits come from?

    0
    #188676

    MBBinWI
    Participant

    What do you mean there’s nothing you can do?  If you’re workload suddenly increases (more forms) you need to be able to react.  wouldn’t it be nice to be alerted so that you can diagnose if this is a permanent shift or just a one-off anomaly?

    0
    #188677

    ?
    Participant

    Control limits have nothing to with specification limits they are totaly set by the process.

    0
    #188678

    TG
    Member

    I don’t really understand the term “passive process”. If that is referring to the fact tht you don’t really control the volume of inputs, isnt that the case with most service processes…you’re not sure how many applications you will get ..phone calls etc.  I’m sure the company knows how many lives are on the books, and from that you can develop a volume expectation, in any case I don’t think you should use it as an excuse to not analyze the data

    0
    #188679

    MrMHead
    Participant

    Not to argue, but to understand:
    If your customer says acceptable limits are +/- Y, that would be your Specification Limits.Your process [ F(x)=y ] output has a variability of z for a given range of input (+/- x)You want your production output to remain within the spec limits (+/-Y):
    F(x’) = Y – z , F(x”) = Y + zSo the input to your process needs to be between x’ and x”
    Aren’t those your control limits?… maybe it’s terminology I’m missing?Thanks

    0
    #188680

    MrMHead
    Participant

    . . maybe I should have used USL and LSL rather than +/- Y for clarity.F(x’) = LSL + z
    F(x”) = USL – z

    0
    #188684

    Marcy
    Participant

    Thank you, Robert, for your thoughtful reply.  We aren’t actuarials so we aren’t trying to trend for peaks and valleys of participant deaths.  We pay if we get 800 or 1600 forms/mo.  A family member may not get around to submitting the necessary forms for weeks — or months — after the death.  It’s a fraught time in most people’s lives.  We can also get multiple forms reporting one death: the wife, the mother, the sibling, etc.  Your insights are really what I’m trying to do in my job in totality.   Right now I’m just cutting up the elephant a little at a time.

    0
    #188690

    Mikel
    Member

    Marcy,This is simple. The number of forms should be predictable
    (poisson distribution) with possibly some type of predictable
    growth or shrinkage (growth in number of baby boomers getting
    old for example).A control chart will not buy you jack. If your new BB thinks a chart
    is important, you made a bad hire. Cut your losses.What problem are you trying to solve?

    0
    #188699

    MBBinWI
    Participant

    Marcy:  Be careful of the help you get here (like from “helper”).  You get what you pay for.  Think through any guidance and understand what is being advocated.  If it doesn’t make sense, you either need more description, or it’s nonsense.  In this case, calls for “bring it under control” would seem to be non-sensical as you don’t control the incoming.

    0
    #188700

    Marcy
    Participant

    Stan — Oh, that made me laugh!  He wasn’t hired as a BB, but decided to go to BB school after working here a while.  Really — he’s a good guy, very earnest but a little green.  I’m also a BB, but I have always had positions where being a BB was a nice asset to my position; however, I’ve never done straight BB work 100% of the time.  I will look into the poisson distribution.  Thank you for your thoughtful answer.

    0
    #188701

    Helper
    Participant

    Marcy,
    MBBinWI is right.  Be careful of the help you get here.  Just in case the MBBinWI did not understand my point, in order to make improvements you must first bring your processess under “statistical control.” (Deming – Out of the Crisis) That is not Helper’s suggestion it is the instruction from W. Edwards Deming.  The instruction he gave to Toyoda and other Japenese executives in helping them rebuild their industrial capacity after WWII.  It is true, one may not be able to “control” the incoming but you can control the quality and productivity of the thru put and that is what you should measure the stability of; the quality, productivity and capability of how the system functions. 
    This guidance is LSS 101 or remedial learning for a master black belt.

    0
    #188704

    Mikel
    Member

    All right, show of hands for who has dramatically improved a process
    that was not in control.I’ve got two hands and one foot up.BTW – I did make them predictable before I was done, but it wasn’t a
    precondition.Deming is right that a process has to be predictable, but I sure can
    study and learn from it along the way.

    0
    #188705

    Helper
    Participant

    Hello Marcy,
    After reading your second post, I have somewhat of a better understanding of what you are dealing with.  You have a valid point in that a run chart can be as useful as a control chart in your situation.  You can detect shifts, trends, and patterns with each – even if the patterns are not just statistical but also seasonal.  The one advantage of the control limits is the detection of aberrations or outliers like deaths due to natural disasters and not just natural causes.
    Buuuuuut if you are trying to “predict” the variation in applications rather than measure the variation then I believe a more useful tool can be regression analysis.  If y is a function of x or x1 x2 x3 etc….  You can endeavor to figure what deaths are a function of.  For example if the main reason for deaths is natural causes then = median age may become one of your explanatory variables.  To start here is what I would do: 
    1.  Segment your data by regions of the country.
    2,  Determine the median age by region.
    3. Collect your data Y – # of Deaths and X = Median age by region and by month(or some index of age, senior population and total population).
    4.  Ascertain the correlation and determination coefficients for your regression model.  If the coefficients fit well, then use the model as a feedback mechanism to determine if it can be used as a good predictor (y= mx + b + chance).
    Sorry for the dissertation, but I truly believe that births and deaths fall on a normal distribution and can be predicted for a population.  As a consequence, your application should be predictable within certain margins of error.
    Hope this helps

    0
    #188706

    MrMHead
    Participant

    But Marcy isn’t looking to predict death rates.
    Her concern is with the rate of forms arriving to be processed.
    She has already indicated that there is a wide variability of time between the time of death and when the forms are submitted.

    0
    #188717

    Helper
    Participant

    Well her response variable will have to be the number of applications then.  The number of days between death and application can be plotted on a XMR chart for that month and the number of applications per deat can also be plotted on a control chart. for those months.  At the end of the day there is alot you conclude if tendencies reveal themselves.

    0
    #188718

    Mikel
    Member

    Definitely not a normal distribution. Do your homework. It is simple, some early life failures due to special cause and
    manufacturing defects (genetics), and a wear out period. Classic
    reliability problem.The forms per x will follow a Poisson with the exception of some
    special catastrophic event. Trending up or down can be predicted
    years in advance with the exception of the special event.

    0
    #188719

    Mikel
    Member

    What is your hang up with a control chart? Identify the distribution
    because what is happening is predictable. Knowing the distribution is
    infinitely more valuable than 1000 control charts.

    0
    #188723

    Severino
    Participant

    You’re just plain wrong.  Control charts are designed to compare the way the process is behaving right now compared to how it has historically performed.  It has nothing to do with specifications. 

    0
    #188724

    Severino
    Participant

    I disagree.  Assume she is attempting to improve the turn around time as a project.  She trials a few different forms, processes, whatever through DoE and finds what appears to be an optimal solution.  She then enters a longer verification phase, but suddenly finds she fails to improve the TAT.  If she had a process behavior chart for the amount received each day and she found that the number of requests entering the system had significantly increased this may impact the conclusion of her study.  This is just one scenario in which the process behavior chart would be useful in this instance, there are others…

    0
    #188726

    Helper
    Participant

    Thank you Stan.  You are smarter than I am. 

    0
    #188727

    Helper
    Participant

    Thank you Stan.  You are smarter than I am….

    0
    #188730

    Mikel
    Member

    At least more experienced

    0
    #188731

    Mikel
    Member

    Wrong.A behavior chart does not help her describe the underlying
    distribution.If she comprehends the distribution, she will not make a mistake on
    her solution.Duh.

    0
    #188735

    MrMHead
    Participant

    OK – I’m just plain wrong – I’m asking for clarification here.Is there not some kind of “Thing” to watch what is going into your process so you can be fairly certain that it will perform successfully?If you’re just watching what is coming out of a process – and you see a defect – it’s too late. Wouldn’t it be nice to have some defined Limits or Parameters for the input so you can be comfortable knowing that 99.999% of your output will meet your customer specifications. And if your input strays outside of those defined limits, by chance you have defined a contingency plan?And from your answer, a control chart would be useless for Marcy because the arriving forms have no bearing on the way the process is behaving right now.

    0
    #188736

    MBBinWI
    Participant

    Darn good thing that wasn’t two feet!  I’d have had to rat you out to the boss.
    BTW – you need to be disqualified from most prolific – you have an unfair advantage.

    0
    #188739

    Mikel
    Member

    Control Charts are way over emphasized in most Six Sigma
    courses.It’s I have a hammer, the world looks like a nail.First, we still don’t have clue what problem Marcy’s associate is
    trying to solve.Second, this is a queueing problem, perhaps we should think
    queueing tools, not control charts.The forms are an arrival rate, it will be best understood in that
    context.I agree with where you are trying to go, control charts are an
    unneeded diversion.

    0
    #188741

    Mikel
    Member

    I’m pretty sure there has been an emeritus award created just for me.Something like – Most Ornery, Most Prolific, and Most Helpful to those
    who really want to learn instead of Spout Dogma EmeritusKind of rolls off the tongue, don’t you think?

    0
    #188743

    Helper
    Participant

    And obviously more knowledgable….

    0
    #188744

    Mikel
    Member

    Correct

    0
    #188766

    MBBinWI
    Participant

    only after prolific quantities of Grappa.

    0
    #188780

    Severino
    Participant

    So you are suggesting that measures of central tendency do not change, variance does not increase or decrease and that I need only understand a distribution and not the time sequence of the data from which it was developed?  I never suggested that a control chart would be the only tool I would use, but I still think it has some value even in this situation.  We can agree to disagree it appears.

    0
    #188784

    Mikel
    Member

    You are wrong, disagree all you want.The lady has a stable system of arrivals, she needs to learn to
    describe it and set up processes to accommodate it.

    0
    #188786

    Mikel
    Member

    Jsev,Before you take it any further, this is a queueing problem. Find me a
    reference to control charts as part of understanding a system of
    arrivals and service and I’ll admit I’m wrong.But don’t waste too much time, you will not find such a reference.

    0
    #188787

    Andrew Banks
    Participant

    Marcy:From what I can see, there’s some good advice
    scattered amongst the rest impugning each other’s
    skill and knowledge. Any time I look at data for
    the first time, I do a few things: plot a time
    series, plot a histogram, calculate some descriptive
    statistics for center and spread (mean, median, Std
    Dev, quartiles). What does this preliminary
    analysis tell you? Are there trends or patterns in
    the time data? Is the frequency data
    (histogram)symmetric? Are the mean & median close
    or not? If the mean & median are similar, try a
    test for normality. Is the data well represented by
    a normal distribution? If yes, then the process can
    be adequately described by the sample mean & std dev
    (i.e. the 95% of the data is between the mean +/- 2
    std dev). If not, then maybe a box plot is more
    appropriate, where you use the median & IQR (you can
    calculate any percentile if you wish).About a chart to represent this process: a time
    series is OK, but if the data is well represented by
    a normal and has no “special cause” variation, you
    could look at an individuals chart of forms received
    per day. The advantage of the control chart IS the
    UCL & LCL. They are “sort of” the +/- 3 std dev
    marks. The idea is that even though you have no
    influence on the arrival rate, you want to be able
    to determine how stable the rate has been over time.
    If indeed it is stable over time, you can safely
    “predict” that the rate in the future will be
    located between the LCL and UCL unless a special
    cause is acting on the process. If this happens, a
    control chart will signal this and you can
    investigate the signal and respond accordingly. Is
    the change a single point,a shift or a trend?
    Single points may not warrant a reaction, but a
    shift or trend in the input may require a change in
    your company’s ability to handle it (i.e. labor). Taking this to the next step and presuming that your
    boss wants to be able to meet the incoming demand x%
    of the time, you could use either the descriptive
    statistics (center & spread) OR a control chart to
    display what the “demand” has been (and will be
    unless something changes).Then do the same for your company’s “productivity”.
    How many forms complete the process each day? Each
    week? Do the center and spread of the data line up
    (input/output), or are they statistically and
    practically different? Maybe run a 2-sample T-test
    on the 2 groups: reason to believe they are
    different? Then maybe there is a “problem”: time to
    find a sponsor(s), charter a project and select your
    team.Good Luck,

    0
    #188789

    RonJ
    Member

    Marcy, as I see it, a Process Behavior Chart will tell you your Natural Process Limits. Assuming there is only common cause variation, your work area should be prepared to handle a number of forms up to the Upper Natural process Limit.

    0
    #188792

    LZ
    Participant

    I think that you have part of your answer already – run chart for the occurrence and work with the Supervisor what key metrics (if established/or if not help get there) is a meaningful measure.  You can’t control the occurrence new arrivals, but are there specific processing (e.g., tat, key steps in process, etc) that would be good to measure with Control charts to show the internal process is in Control.  Sounds like part is determing what you want to measure vs “just count” to track for changes in frequency.

    0
    #188793

    Mikel
    Member

    BBinNC or should I say tool zombie?This is a QUEUEING problem. Her data will not be normal. DMAIC
    doesn’t teach that every problem is the same. It doesn’t teach to
    blindly throw tools at a problem. It teaches -Define the business problem you are trying to solve
    Describe the existing process
    Describe what goes wrong
    Verify you know how to measure (fix if needed)
    Describe the current process behavior with numbers and pictures
    Use the right tools

    0
    #188795

    Andrew Banks
    Participant

    Stan:For the record, I never mentioned DMAIC or LSS until
    the end of my post. The post does NOT indicate
    there is a business problem. What was stated was
    the desire to better understand the arrival rate.
    What better way to do this than some descriptive
    statistics on the data(center & spread), both in
    aggregate and over time?Tool zombie? Time series, histogram, I-chart. I
    count 3. All of which have a defined purpose in
    understanding the process mentioned in the post.
    Hardly done without thought or reason. And
    certainly not related to the eating of brains…Please explain the need to impugn?Regards,

    0
    #188796

    Mikel
    Member

    Can you say QUEUEING?

    0
    #188799

    Jonathon Andell
    Participant

    I agree with you that displaying the data through time is appropriate, especially if it goes back a few years. And control limits may not provide anything actionable.That being said, and given the ease of using most software packages, I cannot see the harm in plotting the data in several formats: – Runs chart (non-parametric – ignores the amount of change, but looks at the patterns of “ups and downs”) – Control chart (for historical data you may as well invoke all the Western Electric tests) – Maybe even Winter’s Method (good for detecting cycles and long term trends).If one of them tells you something useful and new, great. If not, you should not have lost more than 10 or 15 minutes’ worth of time.Maybe you’ll get information to anticipate how work loads will be affected. Perhaps that could lead to an adjusted “demand” estimate for takt times, or it might guide you for scheduling staffing.If you really need to bring the process under statistical control, the first solution that pops into mind is to go start killing some people (KIDDING!!!).

    0
    #188800

    Marcy
    Participant

    Original poster here again.  Thank you.  I will dust off (literally!) my books and look into regression analysis.   (That was part of BB class where my eyes started to glaze over.)  We’ve never done any research into ages or geographic regions of plan participants who have died, but of course the majority of our reported deaths are for people 60+.   That might be an interesting project; however, I ‘m not sure my bosses would see its utility.  I will noodle on it.  

    0
    #188805

    Marcy
    Participant

    LZ — That is good advice.  Note to self:  Look up what “tat” means…

    0
    #188806

    Marcy
    Participant

    Correct.  We really can’t control the incoming.  We have in excess of 3 million participants and they have many beneficiaries.  And for all of you (and I know it’s well-meant advice) who have advocated sweeping research and data collection and number crunching, well, I’m one person here.  LOL!  (And I freely admit I’m not ne plus ultra of BBs.)

    0
    #188807

    Andrew Banks
    Participant

    Turn Around Time

    0
    #188808

    Mikel
    Member

    Jonathon,The incoming process is predictable – I’ll take bets. Any work needs
    to understand that.

    0
    #188812

    Andrew Banks
    Participant

    Stan:Yup: Q-ing.If the input to this process was well-defined and
    stable (or nearly so), I’d agree this data has no
    chance of being well represented by a normal
    distribution since cycle times are bound at zero (or
    some other minimum = theoretical CT) and therefore
    right-skewed. But the process producing this input
    is hardly predictable(on a daily basis)in my mind. More likely it is determined first by the number of
    forms mailed on any given day, and second by the
    type of delivery selected by the sender (std mail,
    Express or 2-3 day, or Overnight). Each one of
    those delivery methods has a CT that may be known
    (std mail is probably not so well defined – as it
    depends on other factors than a guaranteed delivery
    time).So, we have several CTs which can begin on any given
    day (5 of them with higher probability than the
    other 2) and thus overlap. And did I mention
    weekends?…I digress… My point remains the same: analyze the
    data and describe the process using statistics.
    Once you know what the process looks like you can
    determine IF you have a problem (“where we are” vs.
    “where we want to be”). Then you will be better
    informed to select the appropriate course of action.My Warmest Regards to You Stan,

    0
    #188813

    Jonathon Andell
    Participant

    Seems likely indeed. However, predictable could take on a number of forms. My point to Marcy was that control charts may not reveal anything new, but there seems no harm in investing 10-15 minutes to check it out. I think we are much more in agreement than not.

    0
    #188815

    Diorio
    Member

    Hi Marcy,
    I’ve read most of the posts in this discussion but I am a bit puzzled about what it is you are actually trying to solve?  My understanding is as follows:

    The process involves receipt of benefit forms from possibly multiple beneficiaries for the same individual who has died.
    All forms are processed as they come in.
    Of course, there are many statistical and actuarial analyses you could perform on your data to assess whether or not your population of dying people is generally representative of the population as a whole but at the end of the day, you are only one person and don’t have much, if any, time to do such analyses.
    I agree with most of the posters that a control chart would have limited predictive value in this case and assignable causes for out of control conditions are relatively not actionable unless there was a catastrophic event in a region where your participants happened to live.  But even with that – the chart would only state the obvious.  However – what it would tell you is the fluctuation of your process over time and through a capability analysis – provide an estimated natural specification and frequency for the process based on +/- 3 standard deviations around the mean.  If you don’t already have this information – it may be a useful exercise, especially if the claims per week or month are already in an Excel file somewhere.  It would be a piece of cake to analyze the data in Mintab or JMP.  This may be where your new BB wanted to go with the desire for a CC.
    You indicated forms get processed whether there are 800 or 1600 in a month.  One of the posters asked if you know what your maximum capacity to process claims really is.  I agree this would be a good thing to know.  If you then compared the maximum seen to your capacity, you could see where you stand for workload, manpower planning, etc. 
    So, in your situation and based on what I have gleaned from the posts, what I would focus any improvement efforts on would be how long does it take to process the forms?  I believe another poster referenced TAT (turn around time).  I think from a KPI (Key Performance Indicator) perspective and a customer satisfaction perspective – that is possibly what you should be focusing on.  I’m sure your supervisors would support work in this area.
    I would also focus on lean techniques to help with efficiency.
    What is your process to deal with the duplicate form submissions for the same person?  Hopefully, your process allows you to catch the duplicity in step one so you don’t waste resources double-processing any claims.
    Well, there is much more that could be said but I believe this gets the main point across.  Please advise if this is helpful.
    Rich
      

    0
    #188816

    Marcy,If your process is similar to a production process, you might be able to use what I call an ‘Availability Performance Chart.’ (You should be careful as this is not a standard Six Sigma method. it may even be nonesense :-) But it worked very well for us at Linx in the UK. You know that when the purchasing department starts using so-called Control charts. (Actually, I don’t like the term control chart as I was taught to use Shewhart Charts to find souces of variation, by changing the subgroup size, formation, etc.)The way I tackled a similar problem was to first define my capacity – the number of items (forms) I can reasonably process given my resources hour by hour.let’s say for example I can manufacture 60 printheads a day. I could then construct an np chart and plot my losses, noting any loss of capacity – for example, through meetings, exception problems, etc., etc.If you use an np chart, you might be able to determine if your losses are common or special causes of variation.Good luck,
    Andy

    0
    #188822

    Helper
    Participant

    The value in using regression in this case provides a greater ability to predict applications.  If you have the potential of doing so, within a margin of error, you can prepare your organization’s resources more effectively.  Let me know if you need any help with regression. I will send you an easy step by step worksheet.
    Helper

    0
    #188829

    Vallee
    Participant

    Marcy, Within the forest of information, you received a few good trees to
    work with so I am going to avoid the math discussion. I want to
    get back to the forest. What do you really need to measure and why? You can not take one
    piece of the elephant if you do not know what elephant needs help.
    The argument of what chart is pointless without understanding
    why you need it. You do not have a passive procedure because if you did there
    would be no reason for people to work. Heck, you can even control
    your input if you market for outside state forms because you want
    to make money from other states…. wouldn’t that be nice? In my mind (lots of crazy stuff it in) I would want to know and
    measure:1. Peak times for seasonal death.. if any. This would help me plan
    resources and work schedules.2. What defects occur and when they occur. How many times
    forms are handed off between each other for rework? How many
    times forms get handed out on purpose by design? Match this
    data to item 1.3. What is my max capacity and output? Tie this to item 1 and for
    special occurrences like a mass shooting or accident that would
    show a spike not normally seen.4. Man hours required for each form and cross utilization of
    personnel.Why is this important? What does a delay of form completion result
    in? Does overtime have to be increased? What would a cut in
    personnel do to the current process during peak seasons.People talked about process behavior charts but data must come
    from areas of equal opportunity to be put on the same ImR chart. Just some food for thought.HF Chris

    0
    #188840

    Severino
    Participant
    #188841

    Vallee
    Participant

    Wheeler had to make the assumption that “By comparing the
    counts directly, we are essentially assuming that the number
    of vehicle miles driven in Tennessee in each of these 25 years is
    approximately the same.”So he had to ignore that more people continue to move to the
    outskirts of towns as cities develop. He had to ignore that the
    number of cars have increased as the population increased. The
    numbers were not number of traffic deaths per mile, just the
    number of deaths.So for Marci’s question and to use a process behavior chart, can
    you say the population (not the number of deaths) remained
    stable? Were there more births than deaths? if deaths are seasonal
    would the mR also be in control? She could have a process
    behavior chart for each season (major cycle) if needed to be
    trended.

    0
    #188848

    Severino
    Participant

    All models are wrong, some are useful.  Wheeler readily admits to a lack of complete information in the article.  That aside the only point I’m trying to make is that there are applications of “arrivals” (and also departures) where a control chart is useful.  I’ll give you a practical example besides Wheeler’s own:
    My incoming inspection department (don’t get me started) has a need to maintain pace with production.  Historically about 30 lots of raw material arrive a day.  Therefore in order to keep the material in quarantine from growing and the overall inspection lead time from increasing my inspection department must release about 30 lots per day.  I am not in control of how many lots arrive each day and some days a sea container arrives significantly increasing the number of lots.  I am also not in control of what arrives and some inspections are more time consuming than others.
    The issue has always been, when do we put in overtime because we are falling behind.  Ostensibly I could impose the rule “everytime a sea container arrives put in overtime,” but in actuality this is not always necessary if the other forms of arrivals are low or the current inspections are easy therefore I would be wasing money. A similar scenario applies when someone takes a vacation day or calls in sick.  How do I manage the situation?  I have a control chart for lots received and a control chart for lots released.  The control charts are sensitive enough to determine when a shift in my arrivals has occured or when my output has diminished.  (I actually could have done this by establishing a control chart for the queue, but there are political reasons I maintain the charts as they are).  We have a formally defined action plan which determines when overtime applies and how much overtime applies.
    Since instituting the charts and the reaction plan, I have eliminated the need for my company’s “hot list” and pointless hours spent expediting material.  I have also eliminated the need to argue with my bosses about the cost of overtime (although I have projects in the works to eliminate the need for incoming inspection altogether).  Recently, my company instituted electronic inspection records and I was immediately able to identify a sustained shift in the mean which I am using to create the argument for more resources.
    The point is that I have a direct application in which I am utilizing control charts for a queuing system.  Is it the simplest, sexiest solution that was ever developed?  Far from it, but it does work for us.  It gives me the ability to identify process changes and react accordingly and that’s the only point I was making in my response to the OP. 
    I respect Stan’s opinion greatly and the wealth of his knowledge and experience.  In this particular instance I happen to disagree with his position.  You can all stone me for being a heretic.

    0
    #188850

    Mikel
    Member

    Jsev,The issue is you are speaking theoretically. I’ve done this more times
    than you’ve thought about.Disagree all you want.

    0
    #188860

    Vallee
    Participant

    Read a little closer and you will see that I did not say that you can
    not use a process behavior chart (ImR) for arrivals and departures.
    What I do say is that you probably need to keep different PBC’s for
    distinct departures and arrivals if they are extremely different. In
    your example provided it is okay to have such a chart, In Marci’s I
    do not know without looking at her data.I do know however, that I will not make an assumption (like cars
    driving the same miles every year and ignore the number of cars
    included in the death rate population availability like your first
    article did). Even in Wheeler’s book he tells you to use the ImR only
    if based on meeting the opportunities rule ( I know he wrote the
    article and admitted his contradiction but you either remain
    consistent in your rules or don’t use any). This is not the same
    thing as having a debate about what the distribution is. Control
    charts as taught focus on what type of data it is and how it is
    distributed and how many rules you use to specify a change.
    Wheeler just made it more liberal and less technical. You say you do not have control over the processes but the point is
    we are not talking about control. We are talking about making
    solid business decisions on when and where to use your resources.
    Your charts as described seem to be lagging indicators to justify
    your use of overtime. What leading indicators from production
    would allow you to preplan your overtime? What jobs of
    consistently required more rework. Point is the same. What is the forest you are attempting to
    understand before the tree falls on your head. Preplanning based
    on solid leading indicators is better than justification later.

    0
    #188861

    Severino
    Participant

    Stan –
    I get it.  When it comes to knowledge I’m not even a speck on your shoe.  Hope you enjoy the conference next week.

    0
    #188862

    Severino
    Participant

    I never said the article was gospel, Stan simply asked for a reference.  Even then it was a bit of a stretch.  As far as my own particulars I appreciate your perspective and maintain the same.  I have no real desire to hijack the thread any further than it has already been.  I was only suggesting a counterpoint situation in which it might be useful to maintain a chart for the arrivals.  I also recognize my place within the isixsigma foodchain, so with that I shall shut up. 

    0
    #188863

    Vallee
    Participant

    I have seen posts of yours in the past and have been enlightened. It is
    not a food chain…. maybe a big dog on the porch world with some
    bark and some bite and a little pack mentality. But in the end WE all
    learn, revalidate, or sometimes assume just to have to repost the
    revision. If you don’t think so… try to find an old post about socks and
    probabilities from HF Chris many many many years ago.

    0
    #188864

    pm
    Participant

    Dear Marcy,
    Just trying to put forward my points. Pl. let me know if it helps. Your process, i.e., counting # of forms is the output of some other process (no of death cases) and is dependednt on that process fully, where you donot have any control. Hence, rightly said by you, control charts can not help you to prove stability of your process. Rather, it would define stability of the process (no of death cases per month) before to your process, i.e., of counting the forms. Since no of death cases per month is not fixed, i.e., it’s  varying every month, hence is unstable and draiwng a control chart can not talk much about your process stability or capability.

    0
    #188867

    Mikel
    Member

    You were taught that any number that has variation (i.e. not fixed) is
    unstable?Between that and the dogma you cut and paste, I’d ask for your
    money back!

    0
    #188872

    Mikel
    Member

    Did not say that. I said you want to argue about something you only
    have a theoretical understanding of.Do you know how many teeth an ox has?

    0
Viewing 69 posts - 1 through 69 (of 69 total)

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