Using Six Sigma for Call Center Traffic Documentatio
April 28, 2009 at 7:34 pm #24839
Terplife03Member@Terplife03 Include @Terplife03 in your post and this person will
be notified via email.
I have been recently assigned with researching six sigma and possibly implementing it into one of our processes where I work. Can you give a general explanation of six sigma in lemans terms? I do understand the general concept, but I am not sure that this type of tool can be applied to what I am possibly going to try to use it for. I work at a call center as a Telecommunications Analyst and I run daily reports on call traffic in and out of the building to determine what is the busiest hour of the day, call volume for that hour, total trunk usage ect I have been tasked with somehow implementing six sigma into this process to help plot the busy hours, and to explore/understand the difference between counting and an average, and using Six Sigma to predict if the call traffic activity was in control or not. I have no prior knowledge of six sigma until today and have been researching it frantically. From what I have read, it seems to be something that is used companywide to improve the way a company does business. I have not seen anything that would lead me to believe that I can use it for such a small process as call center traffic documentation. I would GREATLY appreciate any help you could give me as to whether or not this is realistic or not and how I would go about even starting this task.
Thanks.0May 13, 2009 at 4:14 pm #59610
Salomon ThanMember@Salomon-Than Include @Salomon-Than in your post and this person will
be notified via email.
First, I think you are not in the right track of implementing a six sigma effort if you don’t know already what it is. Six sigma requires certification or heavy statistical background.
Now, on to the task. Yes, it is possible to implement a six sigma approach into your assignment, in fact, is part of the most useful analyses you can do because it enables upper management to make capacity plans and other process decisions, specially when it comes to a tech-enabled business process: taking/making calls.
Six sigma (and I will be stating this not quite correctly in order to say it simple) is measuring three standard deviations below a mean and three standard deviations above it. The mean in your case, is de average number of inbound/outbound calls – or traffic, that go through your system in a 30 or 60 minute interval. If you have historical data (namely, same month of last year) you can calculate that average and estimate (based on business growth/shinkage) what the average should be for this month, this year. Now, the standard deviation. Six sigma is based on the assumption that large quantities of process-related outputs or inputs (calls) are distributed Normal (mean,stdev). Which in virtue of some stat theory is true, but in your particular case is not quite. Calls normally follow a different distribution, and it have actually been very much under study for a long time. Inbound calls usualy are modeled by what we call a Poisson Process: and call demand (calls received, in your reports) are usually distributed not Normal, but poisson, and the interarrival time (time between one call and the next at your main trunk) is distributed exponential.
By now, you have figured out that you must know statistics pretty good to understand what you have to do. In simple instructions, what you have to do is find the elements of your Poisson process (avg. interarrival time, average calls per hour, std dev of those), estimate for that particular time interval (say 13:00 on 05/12) the six sigma deviations of those numbers, and compare it to other intervals. If they have fewer calls, then that particular interval is out of standard. Likewise, if there are more than that, then is also out of standard. You will have to reserach what happened in that particular interval to know wether it was something you had the ability to control. Which, for inbound calls is unlikely, because most of the time you can’t control why customers call or when.
For outbound calls, the thing is more complicated. For starting point, the process is different because you DO have control on when and who to call. The process can be modeled using so many different distributions. Grab a book on stats for call centers and follow the author’s suggestions (may even be more than a few). And do the same thing: grab an interval, analyze the data, calculate intervals, investigate and draw conclussions.
Ok, now the hard part. These are theories, and those work pretty well in most cases when the time intervals are homogeneous, that is… the nature of the calls, and the processes laid out to answer and respond to customers don’t change. Which is not your case. Chances are you work in a call center that has a mixed environment: one group (or VDN or whatever vector, skill you use to route the different incoming calls) works on different types of calls than the next, so that means each sub-demand of calls has a distribution of its own and peaks happen at different times during a day. Don’t worry, you don’t have to run this six sigma thing for each if you don’t want. But what you really should do to avoid loss of relevance is to narrow down your study, say to 30 min intervals. That is, every 30 min of “open-to-business” time has its own average and standard deviation. So if you want to study a month, that would be 30 days, times the number of intervals within a day, calculations of six sigma deviations. And you can’t compare one day to the next, or one month to the next. Usually, days relate to themselves: monday to past mondays, weekend to past weekends, next june to last june. So make sure that your estimations and comparisons for, for instance, 9/14/2009 14:00, based on averages taken from all Mondays, between 2 and 2:30 pm, of 2008.
Now, the really hard stuff. Now that you have decided to ask your boss for Six Sigma certification, a degree in statistics (paid for of course) or at least for a grant for some stats books, I must tell you that your processes are not quite what the teachers or authors will say. Your processes are bounded, that is, there is a limit on the number of calls that you can receive or make at any given interval, just because the simple fact that you have a limited number of trunks and carriers. You have to take that into account because it alters the outcome.
However, there is software able to do that for you… and if you are lucky you can ask your ACD/Telecomm IT provider to work in an integration that will allow you to do this analysis without even thinking. However, you must know what you are doing, because you have to input the right data to get right decisions, and the set up of the system and the integration must be done thinking on all these points. So, my strong recommendation is that you get certified in Six Sigma, learn about the great IT/software options that can help you with that, and develop your own Six Sigma model four your problem at hand.
Answering you questions: yes it is realistic, and how you go about it is make sure you understand the task and the tools required. It is not complicaded, but it requires background.
M.Eng. Operations Research
The forum ‘Finance’ is closed to new topics and replies.