Staff Required Formula Help Needed

[Jean]

Our excel spreadsheet has this formula to calculate required FTE.
=+(((Number of Calls*Average Handle Time in seconds)/(3600*0.6))/40)*1.5
Can anyone break this formula down for us?
Thanks,
Jean

[felix]

This is just a guess.
3600 - number of seconds in one hour.
.6 - I would guess this has to do with the expected agent's occupancy.. perhaps related to the Service level.

40 - Likely to be the number of hours worked in a week by an agent.

So I'm guessing you are dividing the total time taken to handle your transactions (AHT * calls) between (minutes in an hour * expected occupancy) and it would equal your FTE need in terms of time (hours).
Its likely that the calls offered and AHT are weekly, therefore you divide between 40 (staffed hours for one agent) and I'm not certain about the 1.5 factor.
this might help, but again. Its just a wild guess.

[Jean]

Thank you Felix.
This spreadsheet was left by a former Ops Manager. Now we "manage" the spreadsheet and recently all our reports are being questioned. The question is sure to come up concerning how we arrive at our Weekly FTE Staff Requirement number.
I can not explain a formula that I do not understand. It would be wonderful if you have an explainable formula I can replace the unexplainable one with.
We are taken to the "woodshed" every week over this spreadsheet. Our meeting for this week is in one hour. (Man, I hate getting yelled at.)
Thanks for your reply. I hope you have a bag of tricks you can share from.
Thanks again, Jean

[felix]

Sure!
No problem.
I hope I'm not too late for your big meeting.

Too figure out your needed FTEs you can use the following formula:

This is if you are considering weekly calls:
FTEs required = ((Calls*AHT(secs))/(3600*40*ExpOccupancy*Shrinkagefactor))*Shrinkage Factor


Where:
Calls - The number of calls you receive in a week. The 40 comes from the number of hours that an agent works a week).

AHT - The AHT expected in seconds which is easier to measure. We divide it by 3600 to transform this times into hours.

Expoccupancy is the expected occupancy that you have for your agents. It might come from the service level agreement that you might have. For example, if you expect your agents to be busy 80% of the time, you would use a .8 occupancy factor.

You might also want to consider a shrinkage factor, where you are considering that perhaps around 90% of your people are going to be where they are supposed to be. For example you might want to have a 10% cushion, in case of abscences or tardies, etc. So you might want to consider the shrinkage factor as .9.
now, this formula that I'm giving to you, is a very easy way to consider the FTEs required, but what i would advise is that you run the total FTEs required based from your forecast, in a half hour basis. Remember that not all calls arrive eavenly, and you have spikes throughout the day. This formula might give a little bit some more information on it, but you might still experience a lot of abandonment at times and some intervals in which you have a lot of idle staff.

I'm not sure if I'm being really clear on this issue. Perhaps you are working in an environment where hours are from 8 to 5 from mo to friday, which might leave you with no need for forecasting.

Regards..

Felix

[Jean]

Felix, you are wonderful. If you ever come down to Memphis look me up. I will take you to Beale Street on the Mississippi River for a night of Blues Muisc and Memphis BBQ.
Thank you for the formula. We will try different factor numbers to decide what will work best for us.
I know the question is coming and now I will be ready with an answer to explain how we came up with the FTE Requirement.
Thanks you so much,
Jean

[felix]

No problem!
Anytime. I don't think I'll be going to Memphis any time soon, but if I do I just might take your word for it!



Just one correction. I bet you already figured it out, but anyway the formula would be:

FTEs required = ((Calls*AHT(secs))/(3600*40*ExpOccupancy*Shrinkagefactor))


Just to make sure you got the right thing.

Brgrds.

[JoeB]

Jean and Felix,

I hate to meddle in your conversation, but I often see this approach to FTE calculations and must say that it is wrong. The assumption behind this approach is that calls arrive one after the other in an orderly fashion. If this were the case, than dividing the load (calls * AHT) by working hours, adjusted for occupancy and shrinkage, would give a reasonably good estimate.

The reality however, is that telephone calls appear in a statistically random fashion known as “Poisson Process”. As we all experience, there may be busy times where multiple calls arrive at the same time and callers have to be queued until an agent is free. During other times, the load may be light and calls can be answered immediately. The critical factor is that there is no way to guarantee that a new call will arrive as soon as the agent has finished handling the previous one.

Because of the statistical nature of call arrival and resolution times, service levels are defined as “x% of the calls answered in Y seconds” and not as “all calls will be answered in Y seconds”, and staffing is calculated to meet or exceed this target.

The article http://www.diagnosticstrategies.com/...c_modeling.htm describes the mathematical model of calls arrival and queuing, and how it is used to calculate staffing. Because the calculations are somewhat complex, most call center analysts and managers use Workforce Management (WFM) software tools to calculate FTEs, forecast performance (service level, ASA, abandoned calls, etc.), and do what-if analysis.

To demonstrate the difference between the back-of-the-envelope and the actual, lets take a call volume of 50 calls per hour (2,000 calls per week) and AHT of 10 minutes (600 seconds).

The proposed formula attempts to calculate the load generated by the calls, adjusted for occupancy rate and shrinkage. To simplify the math we will ignore occupancy and shrinkage for now. Using the proposed formula we have 2,000 * 600 / (3600 * 40) = 8.3 agents, which in reality will be 9. This number can be adjusted to occupancy and shrinkage.

There is no need to mix hours, seconds and weeksÂ…. You can simplify and do the same on an hourly basis: 50 * 600 / 3600 = 8.3.

An easier way to explain this formula is to say that with AHT of 600 seconds, an agent can clear 6 (3600 / 600) calls per hour. To clear 50 calls per hour you need 9 agents (50 / 6 = 8.3)

But this is wrong ! Using a WFM tool like http://www.diagnosticstrategies.com/EasyErlang.htm we calculate the required FTE and find out that in order to achieve a reasonable service level (80/30) we need 12 agents! If you have more stringent service level requirement of 90/20 and wish to reduce abandonment you will need 14 agents.

I hope this helps. Feel free to post any additional questions you may have. And Jean, IÂ’d be happy to help you explain this and have management yell at a mathematical model that has been around for nearly 90 years!

Joe

[DaveA]

Felix
I'm new to the site and don't want to start out on a bad note - I've enjoyed everything I've read so far but........

If you are multiplying the shrinkage factor and the FTE requirement as your formula shows and your shrinkage is 10%, you need to multiply by (1+10%) not 90%. You can likewise divide by .9 (if you are really attached to the 0.9 thing.

If you multiply by 0.9, you are actually decreasing the FTE requirement rather than making allowances for shrinkage.

Dave

[felix]

Dave, hi.
You are completely right. If it were a 10% additional shrinkage factor it should be multiplied by 1.1, instead of .9.

I guess I really didn't look much into it.

joe: Regarding forecasting, you are completely right as well. At our call center, we use erlang C based modeling software to calculate our FTEs requirements, and of course it offers a better forecast than just considering daily calls and number of agents available.

However, as I mentioned before, i wass guessing that this is an environment in which agents have a fixed schedule from 8 to 5, or 9 to 6 and fixed lunch hours, which leave almost no possibility of playing around with schedules.

Now, this formula is very simplistic I agree, but it would offer a first approach to scheduling and calculating staffing needs for a call center.

I can't see much need to create your staffing needs according to erlang C, if you don't have a service level requirement, or if you don't have any room for playing around with schedules.

If you apply erlang C mathematical model for this kind of environment, your requirements might be inflated, and the additional resources might not be justified due to high idle times.
Probably the next steps for an accurate forecast model would be:

Get half or one hour statistics
Develop a Service level agreement
Try erlang C on these half hour intervals with your SLA in mind to get your staff requirements
Make decisions regarding staffing vs. service level, as to how many agents you would need for a half hour interval.

Regards.

[JoeB]

Actually, the two methods are not identical and will yield different results!

Say that we have determined that a staff of 30 can met our service level requirements, and we have measured shrinkage to be 20%. You can calculate the required increase in staffing as:

30 * 1.2 = 36

Or as:

30 / 0.8= 37.5

ThatÂ’s a 1.5 FTE difference! Which is which is correct?

The answer depends on the definitionÂ…. From a pure mathematical standpoint, if 30 agents represent only 80% of the required staffing, then the 100% is 30 / 0.8= 37.5. But if we say that we need to increase current staffing by 20%, then multiplying by 1.2 is correct.

One point of interest here is that we assume (incorrectly) that the additional staff is fully available, where, in reality, they will also be subject to a certain level of shrinkageÂ….

My recommendation is that if you monitor shrinkage and schedule adherence very closely and adjust continuously to needs, then multiplying by the required increase factor should work. If you want to be very safe, divide by the shrinkage factorÂ…

Joe

[DaveA]

I bow down to your math prowess

[felix]

Point taken.

You are completely right.

brgrds

[ChrisH]

JoeB,

While Erlang C is good for determining required staff for an interval, how do you suggest determining required staff based on monthly volume? If you break down the monthly volume to an AVERAGE interval and feed that into an Erlang calc, you will not be prepared to staff during peak times unless you recruit extra hours. On the other hand, if you identify the PEAK volume interval and use that number in an Erlang calc, you would be overstating the requirements as you would have more staff than needed for most intervals.

Is there any one "best" approach, or is it really just preference? If you use average interval volume, do you do so knowing that you will need to recruit additional staff (extra hours) for some intervals and manage down staff during others? If you use peak interval volume, do you do so knowing that you will need to manage down staff (voluntary time off, vacation, and other shrinkage) during most intervals?

What we've historically done, and I've never completely bought into it, is calculate the average interval of volume from a monthly forecast, feed that (and our AHT and SL objective) into an Erlang calculation to determine our OPTIMAL OCCUPANCY. We then use that occupancy in a formula similar to the one detailed earlier by Felix to determine required agents.

Thanks,

Chris

[JoeB]

Chris,

I agree that when the only figure you have is monthly call volume, and you use it to figure out the hourly load, staffing calculations may be way off. Not only will this average miss daily peaks, it will also miss other call volume trends, such as day of the week and long term seasonal effects, if exist.

Once you know the peaks and troughs, you calculate staffing for each and design staffing and rostering according to these dynamics, boosting headcount for peaks, and lowering it for slow periods.

I feel that this is more accurate and more traceable than calculating staffing for an arbitrary average and adjusting the headcount using a fixed shrinkage factor. Among others, it gives you an accurate view of other parameters, such as ASA and occupancy rate, which you do not get by calculating headcount for the average and then adjusting manually.

I hope this helps.

Joe

[Lo Monte]

The Call Center School offers QuickStaff, a free software tool to assist in the calculation of call center resource requirements. It can be downloaded at: http://www.thecallcenterschool.com/quikstaff.html .


KoolToolz offers cc-Modeler Lite, a free Erlang C calculator for call centers & help desks at: http://www.kooltoolz.com/ccm.htm .

I have used both of these & found them to be very useful when forecasting staffing needs.