iSixSigma

How Do You Calculate Spec Limits in Transactional Space When They Aren’t Available?

Six Sigma – iSixSigma Forums General Forums Tools & Templates How Do You Calculate Spec Limits in Transactional Space When They Aren’t Available?

This topic contains 7 replies, has 5 voices, and was last updated by  Andy Clark 3 weeks ago.

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #239149

    Andy Clark
    Participant

    Hi All,

    In the transactional space we rarely see spec limits (USL/LSL) as opposed to manufacturing who are normally concerned with a product’s specification limits – the upper and lower limits imposed on the process. This makes completing process capability analysis tough. Using the variation between LCL/UCL for normal stable data and centering this on the goal of the project is commonly used:

    LSL = Goal – (UCL-LCL)/2
    USL = Goal + (UCL-LCL)/2

    An issue with this is that Cp will be 1 – indicating the process fits within the spec limits. But then again with no spec limits provided it should be acceptable to assume the variation in the process is satisfactory.

    Another method I’ve seen involves setting one of the spec limits based upon a target such as an SLA and then calculating the other so that the goal of a project is equidistant between the spec limits:

    USL = Given via SLA (We was all cases to be below this level)
    LSL = USL – (UCL-LCL)
    Goal = USL – (UCL-LCL)/2

    Similar to the method above This will give Cp = 1 again and assumes the current process variation is satisfactory.

    There is, of course, the thumb in the air approach and some variations of the methods above.

    So, my question is what methods do you use to calculate/derive spec limits when you aren’t provided them? Another question might be what standards do you use for reducing variation in a process? Knowing this could help in calculating spec limits.

    Many thanks in advance…

    2
    #239150

    Chris Seider
    Participant

    The customers (even internal) ought to have an idea of service levels required.

    thanks.

    0
    #239157

    Andy Clark
    Participant

    That’s the problem – often they don’t…

    0
    #239348

    Sergey Glukhov
    Participant

    The problem with customers is that they feel the limits beyond which it becomes irritating for them. Six Sigma team should research this issue an turn customer’s feelings into measurable parameters. Usually a CTQ diagram is used together with brainstorming sessions, focus groups and polls.

    0
    #239401

    Strayer
    Participant

    It sounds like you are conflating specification limits and control limits. Specification limits are set by the customer. They are not calculated. Control limits are calculated from the measured variation within the process. We want control limits to be within specification limits but they are otherwise unrelated.

    0
    #239453

    Mike Carnell
    Participant

    @andyclark I am going to risk alienating you right in the beginning. The problem with transactional people is they see this SS thing as some blue collar process for manufacturing. It has never been that and if transactional people stop trying to figure why it doesn’t apply, learn the concept and then think about it you will find out it does apply. If Transactional people had understood the importance of MSA we would not have had the Sub Prime Mortgage Crisis (note to Mike C I am creating a market for my paper on transactional MSA).

    One of the very first SS projects at Motorola was “time to close the books.” Bob Galvin was the CEO and asked to be able to walk out of his office and say close the books. We wanted the closed globally within 24 hours. Nothing blue collar about that and nothing to do with manufacturing. All you had to do was listen and convert his language into a specification. The upper spec was 24 hours. Not particularly complicated.

    We discovered at Motorola after the first 2 years that improvement got a lot tougher the better you got. Managing to a goal like 6 sigma didn’t make as much sense as managing to a rate of improvement. We were told a 10 fold improvement every 2 years and a 100 fold improvement every 4 years. That translates to a 68% improvement. So you manage your improvement to a rate of improvement. That can be converted to numbers i.e. specs.

    Run VSM’s on your transaction processes. Look at the NVA time. It will be huge. Don’t let people off with goals like a 10% improvement. They can step up tell them to cut it 50%. Our goals at Motorola for the first 4 years was a 50% reduction in Cycle Time per year. People holler and scream maybe to day the whine and cry but this can be done and if I cut cycle time 50% one year after I thought my best shot was 10% I become a more confident employee.

    Creating specs are not difficult. They are simple numerical representations of a businesses expectations.

    1
    #239467

    Strayer
    Participant

    To elaborate on @mike-carnell comment, in the transactional space specification limits are typically set by service level agreements, which are negotiated as compromise between what the customer (external or internal) might accept, and estimated cost to provide. This results in dissatisfied customers while contractors (or internal operations) claim that they’re doing really well and have data to prove it, while customers (external or internal) are dissatisfied. This is not acceptable. Dissatisfaction encourages management to outsource, or find a new contractor, resulting in a new service contract, and the cycle takes another round. I think the bottom line in the transactional space is that good enough to meet specifications is never good enough. Ultimately customers want instant and cost-free transactions. Anything less is an arbitrary compromise. So our goal in this space is continuous improvement and never being satisfied.

    0
    #239468

    Andy Clark
    Participant

    @Straydog – I’m not conflating Spec Limits (VOC) and Control Limits (VOP) I’m simply explaining that in the absense of provided spec limits they can be derived using control limits. It’s one option that can be used but has its flaws.

    @mike-carnell – Some good information – thank-you!

    0
Viewing 8 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic.