Service Level Calculation

Topic Views - 3309

Service Level Calculation
So...... Using every version of the Erlang calc I can find, I'm confused.....
Calls 428
Reporting Period 30 mins
AHT 258 secs
Scheduled agents 60
The requirement for this half hour is c68 FTE but with 60 FTE it's telling me that we would achieve 0% SL. Oddly enough I'm comparing what we did achieve, which was 71%. Assuming that this is the first interval of the day, would we not at least answer 60 calls within 20 secs as we have 60 available agents? Even this assumption would give 14% SL.
Anybody else have such issues and more importantly how do people get round them?

Perhaps your Erlang ...
Perhaps your Erlang Calculator assumed a shrinkage that was higher than your actual percentage, which may cause overstaffing.

So, it might be useful to recalculate shrinkage and use an Erlang Calculator that includes shrinkage in its calculations, like Call Centre Helper's version.

For more on calculating shrinkage, read our article:

Unfortunately, there...
Unfortunately, there was no shrinkage included within the calculation.....


Call Centre Helper

There could be a number of explanations
1. The most likely is that you are pushing occupancy too high. This could seem to drive down numbers in the short term, but could cause other problems (attrition, increases in hold time, handle time etc).
2. Your traffic is spiky. Often if people see a short time peak they will jump onto the phones to help reduce numbers.
3. As you go to lower numbers of staff your abandon rate drives up, so these calls do not get answered. However the calls do not go away they get held for later. You may see different results if you use the Erlang A model rather than the Erlang C model.
4. Your traffic may not fit in with the Erlang model. In this case you may be better off modeling actual behaviour.

You can see a number of these factors in the Online Calculator.

You have not said what service level target, shrinkage, maximum occupancy or average patience values that you are using.

If you are not factoring in shrinkage then you are missing a key factor in the staffing calculation.

There could indeed.....
Thanks Jonty,
To give a bit of context, what I am trying to produce is a comparison to an intra-day forecast and an actual day and trying to provide a mathematical explanation for the reasons for variances in Service Level achieved, ie X% variance because we had more/less calls than forecast, Y% because AHT differed and Z% because the actual heads we had differed from the forecasted net requirement, hence not using shrinkage in my calculations. Thus, the sum of X, Y and Z should equal the variance in forecasted and actual service levels give or take a bit.
1) Reported occupancy for the interval was lower than 80%.
2) I suspect it may be but can't get minute by minute call data from my ACD (due to the scale of the data) to verify this.
3) Agree abandon rate increases, for the interval I am analysing, the abandon rate was c.1%.
To provide a bit of background, I am using standard 80/20, 80% occupancy and our average patience value is over 10 minutes (very much driven by the nature of the business).
4) After spending 20 years in forecasting suggesting that Erlang may not be the be all and end all, I suspect this may be the piece of work which proves it...


Call Centre Helper

Look at the call detail records
The only way to settle this is to look at the call detail records. That will give you more detail.

It could be that your AHT distribution is not following the Erlang distribution. There is more detail in this article.

I suspect that what is happening is when you are in busy periods there is some form of human intervention.

Do you have wallboards publicly showing the number of calls in queue?

A number of human factors come into play when calls spike on the wallboards e.g.
- The supervisors spot it and ask people to jump on the phones (sometimes the supervisors jump on the phones).
- If I am due a comfort break but there are calls in queue I may hang on a bit.
- I may get my lunch break delayed.
- Agents work faster to clear the backlog (AHT drops short term and lengthens when it is quieter).

The only way to tell for sure is to sample the call detail records.

You should be able to pop 428 calls into a spreadsheet and examine them in less than 4 hours.

Want to add a comment?

Not found what you were looking for?

1. Try searching through our site.
2. Still not got an answer?

Why not ask the Call Centre Helper Community? Click here to ask your question