Rolled Throughput Yield

Six Sigma – iSixSigma Forums Old Forums General Rolled Throughput Yield

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
  • #30556


    Looking for advice on the best way to measure rolled throughput yield in a manufacturing environment. Do you measure product or process yield, or both?
    If you are using operators to collect the data how do you ensure its reliable? – Not sure how to conduct an MSA without reducing production time.
    How do you define opportunity for a defect. Is an opportunity each workstation, each operator or each component in a part, or number of parts produced?


    Mike Carnell

    Both. You need to have the data setup so you can stratify it by process and product. You would want to runa hypothesis test between products to determine if different products run differently. If there is no difference between products (accept the null product 1 = product 2 = product 3 etc.) then you can treat it as a common process. If there are differences then you need to understand the issue driven by products (probably design related) and process issues.
    As far as MSA every talley point has its own measurement system. You get data from inspection, test, etc. Every point has a proces that produces data and it has its own MSA.
    The idea of opportunity counts is discussed in another post today. My opinion – run dpu. You may be in one of those exceptions if you are running multiple products on your line and they are significantly different from each other. I had a factory in Texas where I ran anti-lock brake controllers and engine controllers a couple thousand of each per day. No real difference in complexity so I ran dpu. In Arizona we had a central wave solder area where we ran high mix low volume (this was the general environment where most of the original SS consultants came from so you low volume people that say it doesn’t work on low volume – that was where we started. For those that invented SS for low volume – sorry, to late) and there were significant differences in the op counts. It turned out the data was best stratified by designer – some guys just couldn’t design a board that wave soldered well.
    Bottom line don’t do op counts unles you have to. Stick to dpu.
    Good luck.



    Thank you,I’m teaching the Six-Sigma tools to people without statistical back-ground,could you kindly present  a simple example of DOE practical implementation to enhance understanding,as most examples -specially in DOE- are difficult to understand ,specially for beginners.Thanks in advance for your kind cooperation,best regards.



    A classic DOE exercise is the paper helicopter.  You can find more information about it at


    Kim Niles

    Dear VH:
    Regarding rolled throughput, Tom Pyzdek wrote a good article on this at:
    Regarding opportunity counting, there are several ways to do it as follows:

    Use a fixed number (typically 2 or 3) times the bill of material count to get the number of opportunities.
    Use the following formula: Opportunities = C+P+S (number of Connections, Parts, and non-manufacturing Steps associated with the product).
    Look at what the customer considers a reject and counts those as opportunities.
    One additional point to remember is that the determination of opportunities should be tied to quality characteristics that must occur properly in the eyes of the customer.
    I hope that helps.
    KN –

Viewing 5 posts - 1 through 5 (of 5 total)

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