iSixSigma

Non manufacturing Discrete data

Six Sigma – iSixSigma Forums Old Forums General Non manufacturing Discrete data

Viewing 15 posts - 1 through 15 (of 15 total)
  • Author
    Posts
  • #46164

    Paneno
    Participant

    I am new to this and am challenged with a non manufacturing scenario – Reduce frequency and $$ impact ( loss) of change orders in the sell process.  We sell very high end expensive equipment and sometimes ( too often), we experience changes to the order after it has been booked due to either a missed requirement ( expense ) to the install process, a missed line item or an incorrect cost estimate.   Often at that stage, it ends up as a giveaway, thus eroding our margins.
    How would you determine a USL / LSL?    What statistical tools would you throw at this in the Measure phase?    How would you develop a control chart with this scenario? 
    This is not a widget
     

    0
    #152213

    Paneno
    Participant

    Robert,  
    Great thoughts.   I cannot do a pareto yet, as I do not have data by change order, tying to a root cause.     I completed the Define phase with clear objectives, and am working in the Measure phase to get to that subject.   I am first holding a cause and effect brainstorm session to determine the gut feel and experience based potential root causes in the x-functional flow.   I will than take these identified root causes and post retrofit them to a randomly selected number of change orders from last year.   I will than do a pareto, etc.  My dilema is, how do I create a control chart or determine capability.    How do I determine UCL / LCL, or USL / LSL, when the only spec is to not have a change order ( discrete metric).

    0
    #152211

    Robert Butler
    Participant

     
    As stated, your post doesn’t give me the impression that you have much need for any statistical tools.  Rather, I have the impression that you need to do more work over in the define phase.
     You said “sometimes ( too often), we experience changes to the order after it has been booked due to either a missed requirement ( expense ) to the install process, a missed line item or an incorrect cost estimate. ” 
      Question: Is this statement the result of a gut feeling, or your latest weekly report, or the bosses intuition or is it the result of having looked at a meaningul block of past data and developed a pareto chart which has identified these three items as the biggest problems on the list?
      If it is the result of a pareto analysis then how do each of these questions break out?  That is, have you taken the time to try to identify things like the types of missed requirements (e.g. 90% of the time we take the customers word with respect to space requirements only to find their estimates are often wrong), or line items and are there any patterns with respect to these misses and to the incorrect cost estimates. 
      You should also examine issues surrounding the ways in which orders are taken and the ways in which your personnel make sure that they have the customer requirements.
     Do they ask open ended questions? Do they repeatedly feedback what they understand about the requirements both during the time of order placement and afterwards.  What about follow up?  Who’s checking to make sure that what the customer said and what you thought they said are one and the same.  What about feedback before quote? etc.

    0
    #152276

    Chen
    Participant

    Hold on…..you wrote… “I cannot do a pareto yet, as I do not have data by change order, tying to a root cause.”  If you don’t have any data how do you even know there is a problem?  I would hold off on the root cause analysis until I get my hands on some good reliable data.  If you do a root cause now, I am afraid it will only lead to speculation.  Go back to Define and get your data first, then let it lead you one.  Remember, “follow your data.”

    0
    #152274

    Jimmy J
    Participant

    Harvey, Typically when you are working with discrete data you don’t have spec limits or targets.

    When the measurement is discrete, then the definition of a defect is a clearly understood Operational Definition that identifies/describes a defect, such that anyone could understand .

    0
    #152277

    Robert Butler
    Participant

      I’m with Jerry on this one. Cause and effect brainstorming and recording gut feelings about potential root causes is not the way to go when you have unexamined historical data.  If you have just gobs and gobs of data then about the only brainstorming I’d do at this phase would be to ask about potential changes in types and frequency of orders as a function of time of year (and if there have been major changes in what you sell then you will also need to make sure you are picking data from a time frame that reflects your current business). 
       If no one is aware of seasonal effects with respect to order frequency or order type then you and your group need to decide on a meaningful time interval, pull the records for that interval, and bean count error types (and lack thereof).  If there are seasonal differences you will have to give some thought as to representative time periods).
      A summary of this data will identify the big hitters as far as error frequency and error costs are concerned.  Once you know what the real errors are you can start thinking about other things.

    0
    #152278

    Paneno
    Participant

    Thank you all for that input.   My data is poor to start out.   My strategy is to fishbone the subjectively assumed root causes, which everyone has.     I will than take them and retrofit them to around 200 records of change orders. Than I will be able to follow my data and do a pareto or other analysis by root cause.   I think I am on target there.   What I cannot figure out and perhaps I don’t need to, is figuring out how to run a control chart, or calculate process capability.   Also, what other statistical tools would anyone recommend once I have the root cause data on my 200 records?    I am already working on improvements on the reporting system’s capability.

    0
    #152451

    Deanb
    Participant

    Harvey,I would start with the accounting department in the data collection excercise. They may already have an account history, and they may have aggregated these overuns telling you what your last period’s loss function was as a % of sales. Then you can start collecting real data, if needed, using a custom designed checksheet for each new change order. Even before the checksheet data comes in I would consider some obvious simple fixes (with team consensus) such as getting dimensional approvals from customers in writing prior to production, formalizing internal communications, getting a redundant verbal confirmation, increasing the usage of “production holds” until approvals are in hand, and controlling expectations by adding dimensional clauses in the bids, T’s and C’s, and sales calls, etc. When enhanced check sheet data is in hand you can use bigger tools in the tool box. My hunch is by then you will have either kicked the tail off of this problem, or uncovered a monster dragon such as inexperienced new hires, stressed systems (workload), or other systemic adversities. Good Luck.

    0
    #152463

    Jim Shelor
    Participant

    Harvey,
    What we are trying to tell you is you have jumped over the measure phase to the analyze phase in an attempt to decide how to limit what you need to measure.  Projects that start that way are doomed.
    Go do the measure phase.  Stop trying to analyze the problem before you have measured it.  You are highly likely to end up in the wrong place if you try to do a blind analysis like the one you are describing.
    Jim Shelor

    0
    #152465

    Deanb
    Participant

    Harvey,Your project goal should not just be to “reduce the frequency and cost impact” of Change Orders, but to turn a current loss function into a profit function. Most companies make money on most Change Orders.I have worked in engineered capital equipment for over 20 years, and have seen the problems you mentioned. Here are some typical solutions my improvement projects identified and implemented. Your needs will be unique to your situation. For “missed requirements” look at engaging more resources in collecting requirements (such as sales people) and funneling these to a single focal point using simple but formal communications (memos). Often the client tells a salesperson one thing, and sends out the bid package saying another, but expects you to “know.” There needs to be a strong “meeting of the minds” with a customer on requirements prior to bidding. Too many times estimating becomes an island isolated from information beyond the bid package. Bidding large projects requires a team approach.The installation phase is always problematic because there are potentially many unknown impact variables in the field. Try qualifying the bids with a longer list of “excludes” or lists of field assumptions, where an absence of any one may result in extra billable costs. Even listing your field rates for extras can be useful, or state “costed on application.” The key is to set better bid rules, then make sure you have a Contract Management function that enforces the bid rules and gets client sign-offs and invoices out for the Change Orders (#1, #2, #3, etc) on every project immediately as they occur (this is a common failure in many companies).Estimating errors can be reduced by having a highly standardized estimating process to make each bid “inspectable,” and requiring a separate pair of eyes to thoroughly review each bid before finalizing it. This generally increases bid productivity too, enabling more bids to be done more accurately. You can win jobs by erroneously omitting costs, or you can win jobs knowing the costs. I always found the latter to offer more profit potential. Good luck.

    0
    #152469

    Transactional Project
    Member

    This appears to be a straight forward transactional type of problem:
    Problem: Number of change orders in the selling process is too large resulting in lost revenues (lost revenues to be determined during the measure/analyze phase?).
    Defect: a change order that results in a loss of revenue. Your proportion defect is calculated as the ratio of all change orders/total orders in a given time period (probably a p or np chart).
    Outcome measure (big Y to be monitored): # of change orders that result in a loss in revenue (if you have few data originally go for a run chart, then move to an attribute chart appropriate for your application).
    (Preliminary) Key stratification factors: Missed requirements, missed line item, incorrect cost estimate (this list may become extended as you perform your FMEA, see below).
    Key deliverables for measure/analyz phase: SIPOC, calculation of current defect rate, detailed process map, FMEA analysis of input and relative risk to process output, Pareto chart of key stratification factors and Pareto chart of inputs with high RPN on FMEA. You can also translate the key “causes” of the FMEA into a Fishbone diagram.
    Don’t get hung up on the differences between measure/analyze phase. They were originally designed to be iterative and potentially circular (you will focus your problem statement as you gain more information). Now, it’s become a ritual exercise to separate these two interrelated activities. Also, after you have reviewed the process with these relatively simple tools, you may end up with a more focused problem statement (as mentioned above). Don’t get hung up on the control chart and the USL/LSL. Instead focus on the process and how the process generates the defects. The differences between common cause and special cause variation don’t make much sense in many transactional processes. By contrast, the impact of “defects in customer input” are of critical concern. That is why the FMEA is so much more important than a pretty looking control chart. I think you have a focused enough problem statement to start the journey. Good luck!  

    0
    #152474

    Deanb
    Participant

    Your initial subjective review will probably give you the “Where” but not enough detail to give you the true root causes. Like SPC: it tells you where and when an out of control point occurs, then you have to go walking around to find out what really happened and why.I would get the accounting department involved to cost out each “Where” to support a Pareto view of the loss function. Then target the leading “Where” by forming an FMEA team with highly knowledgeable process players in that space. Each “Where” FMEA will likely point to a leading 1 or 2 causes you can focus on. However, if this seems too easy, you are probably right. After drilling down on several “Wheres” you will probably find some common causes that are not so simple to fix, making it clear your Change Order problem is merely the tip of other iceburgs. I have generally found this kind of improvement initiative quickly results in the awareness of a needed comprehensive solution with many phases. Just producing this awareness itself however can be very valuable to management. Be careful of stepping on powerful toes. Take it slow and easy, maintain consensus, and you should find this project has great potential for you and your company.

    0
    #152476

    Transactional Process
    Member

    let’s stay on the ground and call your infamous loss function what it is in this context = lost revenues and the where of the FMEA the process steps, or more importantly the input into the process step. Also to give you another reality check: you don’t get just the accounting function involved to “cost out each “Where” to support a Pareto view of the loss function”. Come back to earth my friend!

    0
    #152480

    Deanb
    Participant

    We are talking about understanding non-revenue Change Orders which are currently being internally costed and posted. Harvey may have this info already, he may not. If not, he needs to get it. Not just production’s estimates, but accounting’s postings which are gospel.Also, how these are internally costed is germain to understanding the economics of this project. Are they costed on direct cost basis, or do they get some overheads and indirects assigned to them, as they should? Are these harming other business factors? What are these Change Orders really costing the company? I find accountants easy to talk to, and at times essential to talk to. Perhaps you do not.My reality check to you-friend: on this kind of project if you do not involve the CFO or their department’s representative, you are just blowin in the wind.

    0
    #152481

    Transactional Processing
    Member

    This is definitely a much more succinct rationale.

    0
Viewing 15 posts - 1 through 15 (of 15 total)

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