Every call centre has them; bad days.
Call volumes are higher than expected and staff don’t show up. The result is a mile long queue of frustrated customers; a group of deeply fed up agents and a set of supervisors running around futilely trying to “manage” the situation.
The post-mortem
If the bad day wasn’t enough, then you get the post-mortem, the inquisition into why you had such a bad day.
The answer is always the same:
- The operations guys will point at the forecast and say “the forecast was wrong, I can’t be held responsible for performance if the forecast is wrong.”
- In return the forecasting guys (guys who have their heads so far buried in spreadsheets that you can only see the tops of their glasses), mutter something darkly about regression coefficients and poisson distributions that nobody really understands and then sink back further into their calculations.
The only outcome is that relationships become more strained than they already were. Then the world spins relentlessly on towards the next bad day.
The secret the forecasters won’t tell you
As a man who has done his fair share of forecasting in the past let me tell you a little secret:
The forecast is always wrong.
If your forecasters were the sort of people who could precisely nail the outcome of a random, chaotic, complex system, day in and day out, then they wouldn’t be forecasting your call centre. They would be playing the euro-millions lottery for 5 minutes every Friday and spending the rest of the time cruising the shores of the Mediterranean in a Ferrari.
Try as they might they will never get it consistently right.
100% accuracy won’t solve your problem
But what if your forecasters were always 100% right, could you answer all the calls then? Would you be able to get all the people into the right place at the right time to answer the calls? Or is customer demand so volatile that you are on a hiding to nothing?
Demand volatility
There is a simple way to see how volatile the demand is into your call centre. Look at the data. Take a year’s worth of weekly actual call volumes and plot it out. Does your demand look like:
For the operations analysts among you (or the statistically needy) it is possible to calculate a measure of volatility (if you would rather not know feel free to skip to the next section):
- Work out the average volume
- Calculate the standard deviation
- Divide the standard deviation by the average and show the result as a %
You now have a magic volatility number. If you feel the need you can benchmark it.
But so what? What does that tell you? The bigger the % age the more volatile the system. The more volatile the system the more likely you are to have a bad day.
How does that help with the bad days?
If you want to avoid the bad days there are only two levers you can pull:
- Work out why your demand is so volatile, and do something to smooth it out.
- Look at your staffing patterns and contracts and do something to make them more flexible.
You will notice that blaming one another doesn’t figure too high on the list above.
Useful post? Click here to have them delivered direct to your inbox
Read another opinion
Image by sun dazed
Hi James,
Reading your post there was one statement that stood out for me: “The forecast is always wrong.”
Having been a bot of an economic forecaster in a previous life, I would suggest that your statement is true but also that it is not true and that the forecast is never wrong and can never be wrong. In actual fact, as all forecasts are based on a model and that model is based on assumptions and data, the forecast can never be wrong. Only the model’s mechanics and the assumptions and data that it uses can be wrong.
Or, are we saying the same thing?
Adrian
Hello James,
Excellent post. Yes, the forecast is always wrong. I am into systems thinking and your situation reminds me of Ashby’s Law of Requisite Variety that goes something like this:
“In active regulation only variety can destroy variety. It leads to the somewhat counterintuitive observation that the regulator must have a sufficiently large variety of actions in order to ensure a sufficiently small variety of outcomes in the essential variables”.
Which again goes with your point about having flexibility of staffing options to respond to the variety of demand. This is also where it help to have multiple channels at your disposal to provide customer service including self-service via digital channels.
The real leverage is in reducing the variety in the demand in itself. How much of the demand is generated by the call-centre itself by not providing the right answer. Or the organisation not doing what it had promised to do and so driving calls into the centre. John Seddon has built a business by taking a good hard look at the nature of demand and figuring out what the value demand is and what the non-value demand is. In the process he has saved his clients many millions
All the best
maz
Thank you for the lead Maz, I will go and have a read
James
Hi Everyone,
I want to learn how to create/prepare/generate forecast step by step. Can some guide me on how to do it or does anyone have any tutorial from where i can learn. Please share.
Thanks Prem
+91 9920237722
Prem,
First work out the customer demand.
Look at average calls per customer per month to give you a baseline, then multiply that by the number of customers to predict calls per month.
Then look for repeating call arrival patterns, you will see a pattern by day of the week and by time of day. You might also see weekly or monthly seasonal patterns. Multiply the pattern through to give you an adjusted call forecast. Ideally this should be by hour of day as you will staff to that level of detail
Next overlay any special events that drive calls, eg a sales event or a holiday
Finally work out what you average handle time per call is and multiply that through to give you a total customer demand (in hours) that you need to staff to.
Second step is to work out staff supply.
You know how many hours you need to cover so take that as your base working time and add back “wait time” between calls to give you a total “desk time”. (Google – occupancy erlang call center, for more info).
Finally add back “Shrinkage” ie off phone time, holidays, breaks, sickness, training etc to give you the total paid hours you need
Hope that helps. For more info try http://www.planningforum.co.uk/
Good luck
James
James:
You are absoloutley correct that the forecast can never be right. The historical volatility of offered calls in any interval over the past three weeks typically varies by 60% or higher. But that’s not why call centers have long wait times. The real reason they have long wait times is that the forecast is always right. You see WFM vendors figures out a long time ago that Offered Call Forecasts were impossible to work with. So they made up a formula: Offered Calls = Answered Calls + Abandoned Calls.
They use this formula to manufacture a fake call count every 15 minutes. So no matter how many calls are offered, they only count the ones that get answered, plus a few abandons. Its called “Capacity Based Forecasting”. Calls are offered to the queue at the rate they arrive. They Leave the queueu at the rate they are answered plus abandoned. Arriving is the opposite of leaving. When software forecasts with the opposite of demand, the forecast take on the illusion of always being perfeclty accurate.
However, when a business grows and their forecasts are “accurately frozen” to the historical call answering rate, wait times grow to profit crushing proportions. Planners don’t really care about that though becasue wait times have no impact on forecast accuracy.
The dynamics and perrils of capacity based forecasting are explained in a series of explainer videos that can be found on http://www.systemcapacityoptimization.com
Thanks for your comment Paul, But I’m afraid I don’t really get it.
Could you clarify a little for the hard of understanding?
James
Ditto James. This has to be the most confusing thing Ive ever read, I’m inclined to think it’s a spoof.
Thanks Paul, glad it wasn’t just me :)
Especially in small contact centers there is no usage for highly complicated software that forecasts. Microsoft Excel on the other side is not user firendly and takes a lot of time.
Agenses has a solution that imports your phone systems data and creates forecasts in minutes. This can then be your basis. Feel free to take a look at https://ww.agenses.com
Good luck with the plug Paul :)