Sample Size for Avg. Handle Time

Six Sigma – iSixSigma Forums Old Forums General Sample Size for Avg. Handle Time

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
  • #46211


    I was told recently that anytime that relates to time or $ value cannot be sampled. For example, in a call center scenario, if were to look at Handle Time per call and reduce the average, you cannot look at sample..
    Is this the way how others follow, “Anything related to time cannot be sampled?”… please throw some light on this topic..
    Why cant I look at a sample and analyze the reasons for corrective actions rather than looking at the whole population..


    Jim Shelor

    Time and $ are basic measuring variables that are used often.
    The comment may have been specifically targeted to a call center environment.
    You have to be careful if you are trying to reduce the average time of calls in a call center.  The character of the call greatly affects the time of the call.
    For example, some calls are one question followed by a yes/no response while other calls are several questions that are complicated and require some on the spot research to answer.
    Using an average length of a call as your measure for improvement does not do you much good in that type of scenario.
    Jim Shelor


    Sampling unit

    Sampling requires sampling units. The sampling units are the elements of a population. In a call center you have call center reps (population). They can be sampled and measurements of handle times can be recorded. Also, sampling units can be taken over time, which is referred to as time sampling. However, time itself does not lend itself to the logic of sampling.



    Thanks Jim!
    Let me put some specifics to my example for better understanding. If I need to reduce the time taken on call, I will be required to analyze the top 5 / 10 reasons that are contributing to longer time and address them.
    Going by the Pareto principle, In this scenario, the top reasons are going to be repetitive. Why cant the calls be sampled to identify the reasons and corrective actions be taken rather than going for the whole population.
    Imagine a call center with 1500 calls a day, if i were to go by population count, I will end up analyzing all 1500 calls over a month (1500 / Day X 30 Days), which does not justify the time invested.
    I would rather go for a stratified random sampling by shift / day for picking up calls and identify the top reasons and address them.
    What do you think may go wrong in my approach??


    Jim Shelor

    Sampling can be used to effectively gather the data you need.  You could sample:

    The first 50 calls of the day
    5 calls per hour
    I am wondering about is what is the purpose of trying to drive down the average length of the calls?
    There are a number of adverse effects that can occur from arbitrarily trying to drive down the average length of calls:

    Your current long calls are probably being caused by complicated questions that require some research by your associates to answer.  If these people get the idea that keeping the calls shorter is the important thing, they may shorten their research and guess at the answer.  Instant customer dissatisfaction.
    Some calls are being lengthened by customers who are not fully prepared to make the call and do not have all the facts at hand.  Do you want your associates to tell them to call back after they have all the data they need?  Guess how many are going to call back.
    If you are not careful with this, you can drive a lot of fear into the workplace.
    There are any number of things that you want to happen that can be adversely affected by a drive for shorter calls.
    If I were you, I would better define the goal of this project and attack it a different way.  Just to shorten the length of calls sounds arbitrary to me, probably as it does you.  What is the real reason behind this?  If this is indeed an arbitrary call from above you, build a case for why this is or is not a good idea.  If you decide it is a bad idea, talk to them about it.  That is really what they are paying you for.
    Have a great day,
    Jim Shelor



    I work at a Call Centre and i have dealt with similar scenario with the following Q’s :
    Do i have understanding of the type of calls i recieve.Can i list them in the ascending order of the standard AHT to be achieved( if not a list can i atleast group them into categories like ..installation call,pw reset call,complaint call)
    Can i get the Telephony software to give me data on the AHT achieved in these group of calls.(if not i can get the team to classify the calls on an excel sheet)
    Do i see an area for improvement in the AHT achieved in the specific group.
    Do i see a trend in the CSR’s with higher AHT across all groups.Is it a training issue,attitude issue or any other issue.
    Do i need to set proper expectations from my clients/stakeholders or even the end customers so that i am not pressurised to reduce my AHT every single quarter.
    can i use technology(IVR,vocie messages) or any other idea to reduce long handling time for my calls.
    Let me know what you think.



    Hi George
    There is nothing wrong in sampling for AHT. Here is what you should keep in mind
    There are too many variables impacting AHT therefore the variation might be huge with lot of outliers. Try rational subgrouping (i e dividing the day into intervals within which you think nothing much changes w r t type of calls, profile of agents, call volume etc and pick samples) remember that sub group to subgroup there might be changes happening ( new shift coming in, trainee batch hitting the floor, call volume changing etc)
    Plot this data and baseline your AHT. Use mean if data is normal else use median
    The typical projects on AHT try to find out percentage time taken by talk time (something not necessarily under the control of the call center), hold time and wrap time and reduce hold time and wrap time to reduce overall AHT. Other reasons could be knowledge gap (trainees hitting the floor), bad staffing plan etc
    All the best

Viewing 7 posts - 1 through 7 (of 7 total)

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