How do we calculate FTE Requirement(Half-Hourly/Weekly in spreadsheets. Erlang C seems to create overstaffing. Plz. suggest!

Actually asking a question rather than a statement may help!

What is actually happening?

What numbers are you using?

What have you tried already?

and.....

WHY SPECIFICALLY do you think it's overstaffing?

Regards

DaveA

Dave is correct Pemba,

definition of "overstaffing" may vary from case to case. It all depends on "arrival pattern", "SLA to be achieved" and "occupancy". In case you are contracted to achieve SLA on every interval/hour you would always end up employing more FTEs than if you have to achieve SLA on a day level.

Yes, Erlang C does have tendency to calculate higher FTE, try comparing it to FTEs IEX throws.

Issue of "overstaffing" or in other words "lesser occupancy" can be solved by optmising i.e combining processes of same nature to use idle time and increase occupancy.

THIS IS FROM PEMBA LAMA (ADDED BACK) My assumptions regarding the overstaffing are as given below:

Service Level 75%
Service Time 45
Call Forecast 1
AHT 1800 secs

Intervals = 1800 secs (Half-Hourly)

Agent Requirement according to Erlang C = 2 Agents

However; if there is only 1 call arriving for a particular interval of 1800 secs then why would we require more than 1 agent. A single agent should be sufficient to handle a call with 1800 secs AHT and it should lead to 100% SL; however Erlang C requires 2 agents to handle 1 call.

This is just an example; this calculation through Erlang C in overload situations or high volume gives even higher agent requirement. And even more at weekly level.

**The above agent requirement using Erlang C is without shrinkage. Adding up shrinkage will even inflate the requirement.

Pemba,

Interesting, everyone always tries putting one call in and
seeing what happens........

At the low volume end Erlang WILL always say 2 even when only one call
is expected in an interval. Remember there is always a POSSIBILITY of
a call, and, we're dealing with PROBABILITIES not certainties..

Have a search of the past posts here, this has been covered in some detail before!

Just for reference, using Erlang-C against a traditional Poisson distribution
you get a very similar argument that if AHT >= 10% of interval (ie 18sec for
a 1/2hr interval) and call volume =1 they both give a >40% probability of
a second call , if it arrives, OVERLAPPING the first.

As I said, at the lower volume end all models are squirrelly...

Think of it like this, have you ever seen a set of traffic lights allow
one car out? Even when they are adjusted to their minimum?

Regards

DaveA

Dave is correct again Pemba, that is the primary reason of having "command centre" setus and "RTA seups" in call centres.

Thanks Dave/Pamposh for guiding me through this.

First step is to create a shrinkage factor sheet. Enter the following data in Excel and take it out to 48%:

a2 = 100
b2 = 1.00%
c2 = A2*(1-B2)
d2 = A1:F96
e2 = D2*A2
f2 = E2*(1-B2)

a3 = 100
b3 = 1.50%
c3 = A3*(1-B3)
d3 = A3/C3
e3 = D3*A3
f3 = E4*(1-B4)

Drag the second set of formulas down through row 96. You will have the table below created.

Desired Staff Shrinkage Factor Required Staff Net Staff
100 1.00% 99.0 0 0 0
100 1.50% 98.5 1.015228426 102 100
100 2.00% 98.0 1.020408163 102 100
100 2.50% 97.5 1.025641026 103 100
100 3.00% 97.0 1.030927835 103 100
100 3.50% 96.5 1.03626943 104 100
100 4.00% 96.0 1.041666667 104 100
100 4.50% 95.5 1.047120419 105 100
100 5.00% 95.0 1.052631579 105 100
100 5.50% 94.5 1.058201058 106 100
100 6.00% 94.0 1.063829787 106 100
100 6.50% 93.5 1.069518717 107 100
100 7.00% 93.0 1.075268817 108 100
100 7.50% 92.5 1.081081081 108 100
100 8.00% 92.0 1.086956522 109 100
100 8.50% 91.5 1.092896175 109 100
100 9.00% 91.0 1.098901099 110 100
100 9.50% 90.5 1.104972376 110 100
100 10.00% 90.0 1.111111111 111 100
100 10.50% 89.5 1.117318436 112 100


{EDITED: DA:25/06/2010: Reason: Whole Table note required /
Consolidation of posts}

FTE; Monthly Calculation
The table below would be created in Excel columns A-D. The FTE values in the D column are calculated using this forumla:

=(((Agents(0.8,30,($B2/(24*$C2)),180))*((24*$C2)/173.33)*1.234568))

the last number 1.234568 is the shrinkage factor, use the shrinkage factor that corresponds to your shrinkage percent found in the table created by following part 1.

# Calls # Days FTE Req
Jan 264,592 31 116.58
Feb 211,085 28 90.94
Mar 214,883 31 95.39
Apr 183,328 30 82.05
May 223,861 31 100.69
Jun 224,375 30 97.44
Jul 225,556 31 100.69
Aug 227,928 31 100.69
Sep 229,953 30 102.57
Oct 233,847 31 100.69
Nov 234,299 30 102.57
Dec 238,020 31 105.98

FTE, Weekly Calculation

Using Erlang to calculate the FTE below use this formula:

=(((Agents(0.8,30,B2/168,180))*168)/40)*1.234568

Again the last number is your shrinkage factor

#Calls FTE Req
35357 72.59

Play with the math, you should be able to figure out Daily FTE using your shrinkage factor.

Here is the the erlang formula to use to calculate daily FTE taking into account Shrinkage:

=(((Agents(SL,ASA,#Calls/24,AHT))*24)/8)*ShrinkageFactor

Erlang C only considers hourly volumes for calculations, that is the reason why it would double up the FTE req as indicated by Dave aboove.

Almost!

The actual Erlang formulae themselves can have
900 / 1800 / 3600 seconds in the interval as required
for forecasting. The error creeps in at low volume where it will
always forecast two members of staff for ANY interval as
there is always a >0 % probability of a call arriving.

From simulation this error smooths out once the call volume
of staff requirements go over about 10 calls or 3-4 staff. You need
to rememeber these were designed for high volume environments!

Personally, I'm not sure ANY model (WFM / Excel etc..) will
ever be capable of handling a very low demand environment given
that this is really the antethisis of the reason the stuff is
written in the first place :-)

Regards

DaveA