So this is what the former WFMA told me:

1. Determine the avg % of the customer base calling each month for this year so far.

2. Based on marketing's estimate of customer base number for next year, use the above avg % to determine a baseline # of calls estimated for each month next year.

3. Because two of the CCs I support are 24/7, find out what the avg # of calls is per weekday and per weekend day, look at how many weekdays and weekend days there will be each month you're forecasting for, and adjust the baseline # of calls up or down if there are more or less weekdays or weekend days compared to the same month this year. (I'm not sure if this is right...feedback appreciated.)

4. Look at seasonal patterns and expected fluctuations due to promotions or planned reduction in calls due to whatever projects are coming up and adjust accordingly.

Then we take that estimated magic total number of calls for each month and work with the distribution patterns to guess at how many calls will come in on each day, making adjustments for the days around the stat holidays.

Does this all sound right to y'all? :)

Jen,

Ya got it!

That's about right. However there's stuff you can't anually forecast
for.

i) Changes made to your procedures that have affected call volume.

ii) Call Handle times (AHT) changing driving FTE requirements.

iii) The New Product you launch that generates x% more volume.


Good Luck

DaveA

...to adjust your historical data to smooth out any unique piques / troughs such as promotional campaigns (peaks) football world cup (troughs) etc.

Dylan,

I knew there were bit's I'd forgotten!

I'm sure sacrificing a chicken features in there somewhere too!

DaveA

long term forecasting always seems to end up like that old childrens game...

pick a number (n), -5, +10, -(7-2), /n, +6, x3 = 21 cph if the month has an odd number of days!

Of course you will be needing to take into account exactly which 30 or 15 minutes interval of the day the calls will be arriving.

You will also need to include breaks, sickness (because it's an agents perogative), maternity (they always come in 5s!), and all the other possible tasks that take time out of phone work.

You would save yourself a lot of wasted time if you become friendly with the guys and gals in the marketing department, as they are the ones that will launch their billion call-generating TV ad, when you have half the staff off ecause you believed it to be quiet.

In summary the more you put into your forecast the more you will get out of it and the more accurate it is likely to be, however there has to be a sensible limit otherwise you will over compensate. You will know what makes sense to put in and what doesn't for your company.

On a lighter note...


Have you considered the Box and Line Method, which we were praising only the other day?

Also the widely accepted "finger in the air" method could be used.

The overall favourite though has to be the stab in the dark approach. (Topicl for Halloween)

Dave, completely right a chicken is involved, although a turkey is an optional replacement at this time of year. (My birthday not Thanksgiving!)

Try this link to the very topic to which Simon refers.

John

Thanks guys,

I can't wait until I can use Aspect eWFM to do this stuff for me - they're just getting the second ACD connected to the middleware server (I think) and soon I'll be using that to forecast (whew).

Any other tips or tricks would be greatly appreciated. As you can probably guess, I'm fairly new to doing things this way (the proper way) and I'm anxious to get comfortable with it, because when my boss (a director) asks me if we're going to need more people or not, I'd like to be able to provide an educated answer rather than a guess. I suppose that will come with time, huh. :)

Jen

Hi Jen

It's been very entertaining looking at the posts in relation to forecasting and I can only concur with what everyone else is saying. Having been involved in a number of 24/7 operations there are a couple of things to look out for.

Your call forecast may be based on the total number of incoming calls (offered) remember to reduce that figure by what you think the abandonment rate will be so that your are aiming the agent resource at the number of calls they will answer. Unless the Aspect system can take this into account.

In a 24/7 operation you may not increase / decrease staff based on the volume of calls, say overnight if you have 5 staff on, this maybe vastly more than required but you need a minimum of 5 for say health and safety reasons.

Lastly look at the percentage of time that agents currently employed spend answering calls as a proportion of their total paid for time (around 55%-65%) you can then quickly give indicitive FTE requirements based on an forecast call total for any period of time.

I'm more than happy to discuss this further, in Toronto if possible!!!!

Eamon

Jen,
I responded to your first question...sorry I just started with this forum....I have read most is not all of the comments and agree with most. However, I think it will be easier for you if you put a little process in place ...on paper to assist you....meaning:

1. What Service Level does your "Director/Organization" want to provide for the customer...number of calls answered in the given period of time...85% answered in 30 seconds.
2. What is you customer's tolerance level for queueing?...this can be driven by your service level objective or your historical results from you PBX emulator which should be able to provide you with % and # of abandons and calls answered withing X number of seconds. If you have various categories of calls/customers - you will need to determine this for each category.
3. Determine the call volume distribution (current)...by 15 or 30 minute interval...your choice..or depending on how your application is setup?
4. Now you get into "shrinkage" this can be a make or breaker. Absenteeism is the main one...you need to get with HR to find the number of hours by month and average for the previous year. The rest is easy...training, vacation...etc...
5. AHT (Average Handle Time)...Talk and Wrap(or Not Ready) This should be easy...pretty sure you have been tracking this already....
6. Occupancy or Agent Utilization....based on a full time equivelent 8.00 hours per day....subtract the breaks...if paid...do not include the lunch if it is not paid....you should know this from training...there is the span hours and the paid hours.....span is 9....paid is 8...usually there are 2 15 minute paid periods...which gives you 7.5 available hours on the phones for servicing your customers ....7.5/8 = 93.75%....you need to determine of the 7.5 hours how much time do you want the agent to be talkig to a customer....AHT to Login Time.
7. Plug this info into the application...it will spit out a scenario based on one week (7 day...Monday - Friday)...you then need to review the data in detail by interval..to determine what shifts you need to create that will support the customer and the business....this should take several runs..unless you are just that good.

Other Comment....to Eamon's comment about excluding the nubmer of abandoned calls. I consulted for a company that housed a Meridian(NORTEL) PBX. The PBX could only send answered calls to TCS (WFM system) they never considered the abandoned calls...this create a situation where they used those volumes to predict required staffing....think about it...doing it this way, you are basically telling the customers that never get through (because you are understaffed)...I don't want to talk to you and I will not staff to talk to you)...basically the number of FTE determines the number of calls that will be answered. I suggest that you first determine the total number of call that you recieve for a three month period...find weekly trend..not including Holidays (call vol will be different), based on wkly trend determine interval average number of calls...then refer to "What type of service level do you want to provide for the customer)...also you may or maynot know this, but there are two ways to calculate service level: 1. consider abandons within your delay factor...which would imply that if your SLA to the customer is 85% in 30..if a customer abandons within the 30 seconds it is considered an abandon. 2. do not consider abandons within the delay factor...this would imply that the customer did not give you enough time to answer based on your SLA (service level agreement).