iSixSigma

Measuring an uncertain event

Six Sigma – iSixSigma Forums Old Forums General Measuring an uncertain event

Viewing 17 posts - 1 through 17 (of 17 total)
  • Author
    Posts
  • #36686

    Anand Mehta
    Participant

    Hello:
    I would like to know whether i can measure an uncertain event (statistically/ in terms of money) that if happened could impact business of the company & raise the customer escalation level.
    If yes then what is the simplest way to measure this event…..(please note this event has not yet happened…but if happened will affect one of companys objectives and thats the reason why i want to take up this GB project). Can i take this as GB project or just as process improvement project. Please help/advise !!!
    Anand
     

    0
    #106485

    mcintosh
    Participant

    Anand,
    I didn’t understand you. Please eloborate your problem/objective.
    Tom.

    0
    #106486

    Anand Mehta
    Participant

    my question is how will i measure an uncertain event of an order which comes in standard mailbox and was neglected by the transcation processing team. Cases such as these are few or rare events. But if this happens in future there will be a loss to company..so i want to prevent the same. Please advise how shud i convince my management. they have asked me the for the data. But as i told u its difficult to get such data as its very rare case.
    PS :The mailbox is standard one operated by all. Mailbox recieves all kinds of mails for eg: orders,quotations,price list etc.
    The word neglect means…not processing the manual order and moving it to some othe status unknowingly.
    **orders mentioned are manual orders(comfax orders recieved thru mail)
     Pls advice-

    0
    #106488

    mcintosh
    Participant

    Anand,
    The “rare” event should be converted into a measurable. It can be in terms of DPMO, PPM ……Once you have established a baseline into a measurable, it is quite easy to attach a cost implications for business loss if no actions are addressed. The mangement understands cost better than DPMO, PPM…..
    Once you know your baseline, target a significant improvement & use SS to make your order handling more robust.
    Hope this helps.

    0
    #106525

    Anand Mehta
    Participant

    Dear Tom : You have asked me to measure this rare event. But the problem is the rare event has not happened till now. I am just preventing it from happening…Also i m very new to six sigma…so can you please help me in quantifying the rarity of the event. Please find the required data: 25000 orders recieved for the month of July of which 8000 are manual orders.and rest are EDI(system has completed).
    In this case the rare event is ..that of the 8000 manual orders only 7998 are processed and 2 orders are lost in the standard system it self. This 2 orders were not processed untill the customer came back again.
    I have the solution with me ready to prevent this rare event. But how do i quantify… when the event has not yet happened. Please help me Tom. I would really appreciate that. Also this is my First GB project and i would like if you could guide me thru the define phase of my project. You can contact me at ([email protected])
    Thanks
    Anand
     

    0
    #106528

    Nilanjan Sarangi
    Participant

    Dear Anand,
    I feel, you can simply calculate the process sigma in broad terms ( 2 defects in 8000 opportunities – that is 250 DPMO and 4.98 Process sigma). Any improvement in process would definitely improve the sigma level.
    Also, I am not clear from your mails whether you are doing it for a existing process or a would be process. In case it is a would be process, you should use DFSS tools.Alternately, use DMAIC tools. Anyway, DO NOT try to force a “known” solution to look like a DMAIC initiative. YOu have to follow the rigor of the tools completely to get the benefits, specially when you are in the “upper realms” of moving ahead from 4.98 sigma.
    Best of luck
    Nilu

    0
    #106529

    Nilanjan Sarangi
    Participant

    Dear Anand,
    I feel, you can simply calculate the process sigma in broad terms ( 2 defects in 8000 opportunities – that is 250 DPMO and 4.98 Process sigma). Any improvement in process would definitely improve the sigma level.
    Also, I am not clear from your mails whether you are doing it for a existing process or a would be process. In case it is a would be process, you should use DFSS tools.Alternately, use DMAIC tools. Anyway, DO NOT try to force a “known” solution to look like a DMAIC initiative. YOu have to follow the rigor of the tools completely to get the benefits, specially when you are in the “upper realms” of moving ahead from 4.98 sigma.
    Best of luck
    Nilu

    0
    #106532

    Anand Mehta
    Participant

    Dear Nilu,
    Thanks for your advise. I will certainly keep that i should follow the DMAIC pattern to statistically get the solution. Nilu..the process is already in place and is religiously followed.  My problem is that i dont know how to quantify the rarity of an event which can happen in future and impact my companys business. The rare event which i am speaking here is when the manual order is not processed at all. Till date there was no such case. So can u tell me how should i quantify.
    Anand

    0
    #106567

    Michael Schlueter
    Participant

    Dear Anand,
    “How to measure 2 lost manual orders out of 8000?” One way is to simulate it. Just like you would test a software program, you can also test a (manual) process. In a test you have full control, i.e. you know that this one has to get lost.
    Which guides us to the question: how exactly is an order lost? What has to happen? What are favorite conditions?
    To answer these questions more easily I would ask myself this analytical one: how do I have to use the current system of manual order processing to guarantee that each and every order is lost? 100% guaranteed loss, by just using what is available (no extras, no obscurities, just what you have in place).
    I can’t give you a solution without analyzing you situation, but I can give you a thinking-direction which will guide you to a strong solution:

    the lost order identifies ITSELF (nothing else is needed).
    Think of how your system of manual order processing works. An order is lost. This (now invisible) order has a new property: it can identify itself as a lost one. Nobody can overlook or ignore this one.
    This is thinking in terms of analogy. Think of a lost child. What does it do? It cries. Nobody can ignore, so it’s found again.
    Today your customer has this informing role (“why didn’t you process my order !?”).
    Tomorrow another available resource in your system performs exactly this function. That’s what you have to achieve.
    In analogy you may think of other systems, which have a similar problem. E.g. air traffic controllers do handle airplanes manually. How do they guarantee that no plane is ever lost? (And how do they know, i.e. how do they measure?)
    Kind regards, Michael Schlueter

    0
    #106568

    Michael Schlueter
    Participant

    The lost order identifies ITSELF (nothing else is needed).
    E.g. you can note the order on (or attach to incoming documents) a sheet with a striking color, like flashy yellow or so. No matter where this sheet resides, it simply can’t be overlooked.
    Perhaps it can be overlooked by the low percentage of color blind people. Perhaps that’s why some sheets have cross hatched dark and light bars on it. Think of the Top Secret documents.
     
    Kind regards, Michael Schlueter

    0
    #106569

    Anand Mehta
    Participant

    Dear Michael,
    Thanks for guiding me thru the thought cycle. Michael…i have a process thru which i can identify whether an order is worked or not…i have mentioned this to my manager. But can you let me know hoe shud i convince my Black belt that my suggestion could be implemented as a GB Project.(well i have yet to do DMAIC). What if no manual orders are lost…everything is processed correctly. Can i not benchmark this solution so that there is no loss to the company…How shud i present the data to my BB if there are no defects (loss of Orders in mail)…before asking his permission to start my GB–Pls guide me as i m very new to the concept of GB–Anand

    0
    #106570

    Michael Schlueter
    Participant

    Dear Anand,
    Thank you for your reply. I think it is important to view this situation from the right perspective.
    What was the answer from your manager?
    One perspective you did mention yourself: how relevant is this process? Will your company lose millions of Euros?
    The other perspective is about rare events themselves. How do you know (in a rare event) that there is no more unwanted event?
    I.e. how do you know the people involved have changed and work 100% defect free now? People always provide the chance for error, or at least, non-conformance. You did observe this defect in the past – why should it be gone? When (under which conditions) will it return?
    You have 2 losses out of 8000, or 250ppm. So, does your BB or GB know “it’s perfect” when he/she identifies 0 losses out of 400 investigated processed orders?
    Probably not. Probably you would strive to investigate 80.000 orders instead. Your BB should be able to give you significance limits when he/she knows for sure (in statistical terms) that the error level now is significantly below 250ppm. (This is another way to state ‘no defects’).
    Rare events have the characteristics that the mean time between events can be very large. That’s another way to observe rareness. E.g. when you did observe 250ppm lost orders a while ago and now have 2 or 3 months without any losses: yes, that’s consistent, isn’t it?
     
    So, in summary, when this is a relevant problem, it should be investigated. I would not regard it as idle effort. To make things more practical think of stressing your system, as I said about tests. When you put the right stress on the manual order processing rare events will turn into frequent events.
    E.g. when you take away 10% or 30% of the people, will it still be ‘virtually defect free’? When you flood them with orders, will all be processed?
    You should also consider your own effect in this matter. I assume you did pay attention tho this problem for a while. So the people involved did recognize ‘Amand is watching us’. Can this explain the virtually perfect status of this system today?
    Hope this helps. Michael

    0
    #106571

    Szentannai
    Member

    Hi Anand,
    I’d suggest a Fault Tree Analysis and an FMEA. You could estimate the individual probabilities of each branch of the tree – and then a probaility of a certain chain of events leading to the fault by taking the product and/or sum along a given path.  You could then check whether your estimate correlates with the measured occurrence rate BTW. Given the different path along the fault tree you could build the FMEA and use it as a guideline to define some projects.
    Regards
    Sandor
     

    0
    #106572

    Anand Mehta
    Participant

    Dear Mr. Michael:
    Thanks for guiding me specially in a way when i mentioned you that i was very new to this concept of six sigma.. Sir you are like a God father to me …really. I would be greatful to you if you could provide me your personal email id so that i could let u know or may be take some help at the different stages of my project..incase my BB/manager approves my charter.
    PS:My manager is going to keep a meeting with me very soon and i m going to explain him the case study of my project. Also my process is very critical and 1 of the companys objective is to process orders correctly in the system.If i loose even 1 order …my company might loose million dollars or Euros in that particular order. So i better benchmark this across the process.
    Regds
    Anand ([email protected])

    0
    #106577

    murali
    Participant

    If you know the solution to the problem(8000 orders), then you cann’t proceed as a GB project. It will an ideas to prevent future failure modes.

    0
    #106585

    Ronald
    Participant

    Anand,
    Here how I would approach it:
    If you do a analysis with cost/benefits associated with the probability of occurrence, it should show the likely or expected outcome. 
    Assume 99.99% of the time the process works well and only costs $100 per year to maintain.  But with a 0.01% chance of failure it would cost the company $1 million.  The likely outcome then is (0.9999*-100)+ (0.0001*-1000000)= -$199/year.  But if your system/solution were more robust and granted 99.999% correctness then the next question then would be how much does the improvement increase or decrease operating costs.  Assume it is still $100 per year; the expected result would be -$109.999 per year.  Thus showing improvement.  If it increased costs to $1000 per year.  The results would be -$1010.  Showing high costs compared to the original state (no improvement).  Maybe an analysis like this would help.
    I agree with the FMEA approach as well.  When dealing with the risk factor of a rare event that is difficult to measure, you could move down a more subjective road (Severity of Failure is High, Occurrence is Low, Detection is however good you determine it is).
    Finally, if you have a solution already in hand, then I would not deem it a Greenbelt project.  It should be more of a justified “just do it” based on cost/benefits.  However, if you have a slew of solutions to pick from, then more of an analysis/piloting/implementation/etc is needed and this would qualify as a Greenbelt project.

    0
    #106606

    Szentannai
    Member

    Hi Lee,
    I’d have a concern about basing the cost/benefit analysis on the observed occurrence rate of 1 / 4000 (I don’t remember the exact number – might have been 1/40 000) because this will be a very imprecise measurement of the “true” occurrence rate. To get a reasonably precise estimate you might have to gather an unreasonable amount of data – and accept some high risks as well.
    I would not advise something that pretends that we know a lot about the process when in fact we are almost completely ignorant of it. I’d prefer to use all we DO know  (i.e. a fault tree analysis) to reach the best possible conclusion.
    regards
    Sandor  

    0
Viewing 17 posts - 1 through 17 (of 17 total)

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