We are currently reviewing our customer experience processes, and as part of that are looking to reduce the IVR times that the customer faces.

We don't currently use any automated solution when dealing with calls, so all calls will be answered by an agent, but we just want to understand the impact on the queues of reducing the IVR time.

Obviously, we know that the calls will reach the skill queue more quickly, and would like to simulate or model what happens, based on various AHT's and call vols.

Anyone any idea how we can do this? Any help would be much appreciated.

Darren,

I've never done this but I can see where you're coming from.

I'll have a think and a play with a whiteboard to see if I can
come up with anything.

Regards

DaveA

Darren

In pure Erlang terms this will have no effect at all. If your callers currently spend 30 seconds in IVR and in future spend 15 seconds - this will only shift your entire traffic load forwards by 15 seconds. Thus the arrival pattern at the desk will remain the same.

Unless you have a large number of in-IVR abandons, you do not have to worry.

Lester

Yes - I'm inclined to agree with Lester.

The only conceivable change is that there'll be less people dropping out if the IVR is quicker, which in theory may result in a slightly higher percentage of the calls recieved making it to an agent.

However, I wouldn't worry about this. In most instances these callers that had dropped out would have called back anyway, so therefore your initial call volume may decrease slightly.

Just out of interest, which site are you at Darren?

Regards,
Darryl

need to take into account some calls will reach the wrong skill queue....

Z

As one of Darren's colleagues (the one who asked for his help!), perhaps I can give you some more information.

At present there is a small handful of options that you can make once in the IVR meaning that, dependant on which route you take, you will join the queue anything between 1m 05s and roughly 3 minutes after entering.

The simplification work will result in ALL customers spending only 20-25 seconds in the IVR (fewer choices, fewer menus).

So, let's say that 20 callers enter during the same minute. The current set-up means they are likely to hit the queue in dribs and drabs. Under the new set-up, they will hit pretty much every 3 seconds.

The most obvious result of this in my mind is that queues are more likely to build up as a result. However, if this calling pattern is repeated every minute over a fifteen minute period, for instance, would either set-up be worse or better than the other?

In fact, given that call arrivals themselves are random in nature, is it possible to do any modelling to prove or disprove either theory?

Any thoughts would be welcome.

In fact, given that call arrivals themselves are random in nature, is it possible to do any modelling to prove or disprove either theory?

I think you're right. One of the problems with traffic forecasting methods such as Erlang is that it assume a regular delivery rate, which never happens.

Personally, I don't think you'll see any net increase in queue length. They'll ebb and flow just like they did before.

Regards,
Darryl

If you've reduced the IVR options, then it's possible that you'll end up with less calls going to the wrong place. This may improve handling times and therefore reduce waiting times.

If this does occur I expect it will be on such a small scale that you won't notice, but it's an interesting thought nonetheless.

Anyway...it's getting near home time.

Is is possible to try it on a small scale? Perhaps only for a few of the calls or for a period of a week or two?

If it is you can use the old PDSA cycle: Plan, Do, Study, Act.

Plan
----
Put together a plan of what you will do including how you will objectively measure the outcome. In this case I suggest a control chart of the queues and inbound calls.

Do
--
Enact the plan on the small scale.

Study
-----
Compare the objective measurements from before and after.

Act
---
Depending on the conclusions of study, either make the changes permanent or go back to the drawing board. Then back to Plan.

PDSA (or PDCA with the C being "Check" instead of "Study") is commonly known as the Deming Cycle (though Deming called it the Shewhart Cycle). It is meant to be gone through again and again to continously improve.

It is really just the scientific method. Come up with a theory. Design an experiment to test it. Decide how to measure it. Do it. Study the results. Change or don't change the theory based on the results. Go round again.

Some may have seen it in the Six Sigma form of DMAIC (Define, Measure, Analyze, Improve, Control).

A nice summary of the process, there, Rob

Z