iSixSigma

Sub-group sampling

Six Sigma – iSixSigma Forums Old Forums General Sub-group sampling

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #34859

    AB
    Participant

    We will be sampling from a high volume call center to undersand the variation in handle time.  I understand the concept of subgrouping for the Y, we will be taking a set number of calls every half hour or hour and averaging them for analysis.  My question is in regards to the X’s.  How should I account for different levels or values the X’s can take on within a call we may sample?  Thinking of all of the potential X’s that we may end up with is starting to muddy the waters in terms of the sampling approach.  Hope this makes sense.  Any help would be appreciated.

    0
    #96685

    Primantara
    Participant

    Hi,
    Can you tell em what is the Project Y. What is the metric you are tring to improve?

    0
    #96696

    AB
    Participant

    Call Handle Time

    0
    #96701

    kps
    Participant

    Hi
    In this case first make a high level Process Map Using SIPOC and then define the Small Y’s and then monitor them individually using control charts. Also, use Correlation and Regression tools to study the effect of small Y’s on Big Y(hold time) 

    0
    #97049

    sandi
    Member

    i think you can group it under complaints/queries, then break it further under existing and potential customers, then sub group under 2 major complaint areas which is what you want to hit as per pareto.
    thanks
    sandi

    0
    #97059

    Bev
    Participant

    I wouldn’t average the call times.  ( most ‘cycle times’ have a skewed distribution plus you get more form the detail than from the average) you may plot each subgroup with a boxplot.  (The median is a better statistic than the average for cycle times).
    I would initially subgoup by large “categories” of X’s based on how the Y can vary over time:  time of day and day of week. dependign on your call volume you may want to sample every hour or every 2 hours.  (I usually start with more subgroups and plot the data immediately and redcue the frequency of subgoups if I don’t see any significant variation from subgoup to subgroup.  again depending on call volume and the difficulty of getting call times, I would sample at least 5 calls per subgroup, preferably 10 no more than 15.  If you sample 5 calls, you can forgo the boxplot and simply use an interval plot (vertical dots for each subgroup).
    Since call time is a complex system (as you are aware) I would also collect info on several broad Xs:  Operator, “complexity” (up to you, think about it.), Problem Statement, Product type (if multiple products), etc.  I think you get the drift.  Then once you are done collecting your intital data* you can “slice & dice” by those categories of Xs to see if any standout (Hypothesis test time).  Don’t try to get too detailed at the start…start broad and then focus in on the 1-2 broad categories that are largest.
    *how long shoudl you sample before doing initial analysis?  depends on how long it takes for you to go from “long call times” to “short call times”.  If it’s call to call then a few days is more than sufficient.  If call times “cluster” in long and short groups, then collect across 3 of the short to long groups…

    0
Viewing 6 posts - 1 through 6 (of 6 total)

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