Segmenting Callers into Clusters

Below is the continuation of a transcript of a Webinar hosted by InetSoft in August 2010 on the topic of "Performance Management in Government." The presenter is Christopher Wren, Principal Consultant at GPM.

Christopher Wren (CR): Let me just show you one more interesting angle on this. So this is another call center that we have been working with, and in this case the theory was that the lower you could bring down average talk time the happier the citizens would be. These are people calling about their benefits, and the belief was currently they are on the phone too long, and no one wants to be on the phone with the state government any longer than they have to be, so we really need to bring this talk time down as low as possible.

Well then we look at the data and as you can see something interesting about it which was that satisfaction was pretty high with short calls, and then you can see satisfaction takes a dip as people stay on the phone past about 2.5 minutes. People start being less and less satisfied. But then something very strange happened, which is at about 3 minutes 12 seconds or so, satisfaction started going up again, and this is a real mystery.

What was this telling them? So we disaggregated the data, or in other words tried to segment the callers into clusters of certain similar characteristics. What you can see here is that there were really two different kinds of people calling. There were what we call the angry calls, or people who had just some quick complaints, something they wanted to be resolved, maybe their address was wrong, either the name was misspelled, whatever it was. There were these quick calls. And so these people who just want to get on the phone and get off the phone.

But there was another group of people that they hadn’t really thought about, yet. And those were others who need more care. These are what we were calling the desperate requests. People who had a longer story and needed to really get their story out. They were more complex. They needed to stay on the phone longer, and with these people you didn’t want to cut them off early. So with these people after about three and a half minutes their satisfaction went up and stayed high even as they stayed on the phone for four minutes.

So really what we are trying to tell you here is that one target is not necessarily right for all situations. De-aggregate your data. Look at the data in a little bit more sophisticated way, and you can start to break it apart and see, well gosh there really are two different groups here so we need two different targets, one target for the kind of people who want to get off the phone quickly, one target for more complicated calls. Different stories require different targets and different performance metrics.

More Resources:

		Report Parameter Retrieval
		Using Tabular Data in a Report
		Using Business Intelligence
		Best BI Tools for CRM