Archive for the ‘Call Centre Guru’ Category

Forecasting and Planning for Call Centres

Friday, October 8th, 2010

I was talking about Forecasting and Planning for Call Centres with a gentleman on the plane earlier this week.  He didn`t understand why his Call Centre is putting so much time and effort for poor results.  After going over their daily results with him, it was clear I needed to post on the topic.  Forecasting and Planning won`t bring the same results every day.

First, let’s start with the fact that the Call Centre management team wants to staff at the minimum required to achieve service levels.   The required service levels good be daily, weekly, monthly, etc.

Most of our Call Centres have a weekly or monthly target.  The Forecasting and Planning would be based on this monthly (yes, you can read “average”) SLA.  It is not optimal for each day or each interval.

Why?  Costs.  Trying to hit 80% TSF each day requires more FTEs, time, effort, employee dissatisfaction, etc  than trying to achieve 80% TSF for the month.

Service Level or TSF calculation… what about abandoned calls?

Wednesday, August 18th, 2010

I often get the same question on how we should calculate the Service Level or TSF.  Should we put the abandoned calls in the equation or not?

To me, the answer is simple:  Yes you should.

The reasoning behind is simple.  If you would have had an available agent at that specific time the call was abandoned, you would have answered the call.  Then, you would have the “sale” attached to it, the AHT, etc.

Unless your Call Centre is driven by very specific burst of calls, for example, a contest to win concert tickets where callers call non-stop until they are answered, the callers more often do not call back within the same 30 minutes.

Weird idea… what about changing the ACD coding?

Wednesday, May 12th, 2010

I was talking with a Call Centre Manager about his TSF target.  The Call Centre is a public Call Centre following Federal laws around TSF target.  They need to achieve a TSF of 80/20/80.  80% of the calls answered in 20 seconds 80% of the time, which means in 80% of the intervals of the day.  They are having a hard time achieving their TSF target.

Then I had a weird idea.  This Federal TSF target, on what type of ACD coding is this in relation with?  We all know the core ACD is programmed to route the oldest waiting call to the next available agent, unless another priority is coming in the mix.  What about an ACD code where, depending on the TSF of the interval, the oldest waiting call is not routed to an available agent?  The oldest waiting call is already lost (for the 80/20), so let’s route a call that waited 19 seconds to the agent?

I know, this is not good Customer Service for the oldest call waiting customer, but we have to agree that the oldest call waiting customer wouldn’t know.  Not even the agents or the Team Leaders.  This “solution” wouldn’t work if there are many calls waiting, for many minutes.  But here and there, this might work.

I wonder if there are some Call Centres with such ACD coding…

Why is it so hard to speak to a real agent?

Wednesday, April 21st, 2010

I would simply start with the fact that the technology behind self-service support is easy to put in place and appreciated by many users.

For this discussion, let’s take a mobile phone provider’s call centre.  When you call their customer service 1-800 number, you get prompted for your 10-digit phone number.

When the phone number is keyed in, a link is done between the mobile company databases (billing, options, etc) to the phone system also known as the IVR. (Interactive Voice Response).  From your phone you can verify your account balance, pay your invoice, add options, etc.

Another self-service medium would be the mobile’s company web site.  From the web site, you can login and access your account information and make changes to your account.

We need to understand that for the mobile company to offer self-service  is cheaper than having real agents answer the calls and often provide the same information as the IVR or the web site.  With self-service, the mobile company doesn’t have to worry about operating hours (the web site and IVR are available 24/7) and they don’t need to worry about the AHT.

The self-service has many benefits, but also a major down side.  Having your customer live on the phone is a business opportunity and should be used as an opportunity.  Your agents need to understand that.

This is one of the reasons why many call centres split their phone queues based on the type of customer.   For example, in the IVR, based on the mobile phone number entered, a decisions is taken based on the monthly charges on the account.  Some call centres would route the call directly to an agent if the customer is spending more than 500$ for example.

Tip:  You want to speak with a live agent as soon as possible?  Do not enter anything in the IVR.  Not even language selection.  In that case, many call centres route the call directly to a live agent based on the regional laws around customer service to provide service to houses with old rotary phones…

One call at a time

Wednesday, March 31st, 2010

We all have seen a “regular” call peak for 1 hour around 2pm.  To get rid of the calls from the phone queue, many agents trick their AHT.  They do the minimum needed with the customer, put the file on the side and take another call. (And eventually finish the files at a later time)  In fact, this tricky method is emptying the phone queue faster, but at what cost?

The solution for this call peak is to properly staff at 2pm.  To be able to staff properly we need an accurate forecast on call volume and AHT for the 2pm interval.  If historically the AHT has been tricked to empty the phone queue, the forecast will be wrong and the proper staffing will not be in place.  To avoid all of this, we need to make sure all the agents are answering the calls the same way they do any other time of day.  This will require a change in the Call Centre behavior, which is not easy.

This behavior will have many benefits.

  • One big objective for a Call Centre is to avoid call backs (incoming or outgoing) and try to complete the activity with the customer with 1 call.   By not rushing the call, there is a greater chance to be able to complete the activity with one call.
  • The customer service (quality) should be better when the agent is not rushing the caller.  For the customers, they have been waiting in the queue for a while.  There might be dozen more waiting but for the customer you have on-line, only this call counts.  Having your customer on the phone with you is a business opportunity.  Make sure you get the best out of it.

Shrinkage… for Call Centre Management

Thursday, February 18th, 2010

Of course we all heard the shrinkage joke in our Call Centres… even with the pooling theory!

The shrinkage is an important value to consider in your Call Centre.  The shrinkage is the percentage of time your agents will not be on the phone answering your customers.  There is planned shrinkage (scheduled breaks, training, scheduled vacations, etc.) and unplanned shrinkage (agent missing in action, etc.)

We could argue on the fact that the shrinkage could always be planned.  I tend to agree.  Absenteeism is a good example.  We can’t be sure on how many agents will be calling in sick next Friday, but we know there will be a few of them.  How can we plan and forecast it?  The same way we do for call volume and AHT, we analyze the historical data on absenteeism.  We do the same exercise for system issues, and any other elements that could keep agents from handling customer calls.

How to skill the agents for the Call Centre?

Friday, February 12th, 2010

You have now set up your call flow for your Call Centre. You have forecasted your volume of calls using your historical data and tweaked it with the real time volume, which may resulted in some overtime if the volume is higher than forecasted or letting the agent go home early if the volume is less than forecasted.

When you forecasted your number of agents needed on the phone, did you take in consideration which agent is taking which type of call?  Could this be a reason why your Service Level forecast is not accurate even if your call volume is?  Yes, let’s look at an example for our Road Side Assistance Call Centre:

Let’s pretend you forecasted 200 calls for your Sales call flow, which meant 10 agents needed based on the model and shrinkage. (Yes, the shrinkage will be part of a future post!)  Also, you forecasted 200 calls for your Assistance needed call flow, which meant 20 agents needed based on the model and shrinkage.  The model forecasted a service level of 80%.

In real life, you achieved 90%.  A few of the reasons might be the distribution of the calls, the AHT, etc.  Let’s again pretend the distribution and the AHT was as forecasted.  What else could’ve happened?

How are your agents skilled in the Call Centre?  Are the 10 Sales agents loggued in only in Sales?  What about the Assistance agents?  This could make a huge difference.  Forecasting based on agents that are multi-skilled is complicated when you create your model by hand.  Many Forecasting and Planning application take this into consideration.  But again, are you often changing the agent skills during the day?

What is a skill or skill level?

Let’s take the Sales call flow and pretend there is no option within the call flow.  We would then have one Sales skill.  Idem for Assistance.  Within your phone system skill interface, you would update your agents’ profiles to match their skill (knowledge, training) with the proper Call Centre skill.  Is agent John Smith a Sales agent or an Assistance agent.  In John’s profile, you would then assign the proper skill.  In this case Sales.  Most phone systems would also allow you to assign the skill level.  A skill level is a priority you assign to a skill.  For our phone system, the priority range would be from 1 to 16.  1 being the highest level, meaning the first one.  I recommend using 5 as the default.

Now, let’s talk about Julia.  Julia is an experienced agent that have been in the Call Centre for many years and assigned to the  Sales calls but has enough knowledge to help with Assistance calls.  In her profile, we would then assign her the Sales skill and the Assistance skill because we want to use all the help possible.  For Sales, we would use level 5 as default.  Now for the Assistance skill, do we want Julia to take both type of calls at the same priority or we want Julia to first focus of Sales calls and then, if available, on Assistance calls?  For our example, we would like Julia to first take the Sales calls and then the Assistance calls if there are Assistance calls waiting and no core Assistance agents available.  In this case we would assign her the Assistance skill at a priority higher than 5.  I used to use 12 for backup skills in the past.

Could we use level 1 and level 2?  Yes, of course. In our example, it would make a difference.  But, the example and the call flows are quite basic.  In your Call Centre, I am sure you have way more skills like, VIP, French, English, Spanish, etc.  If your default priority is 1 and you want to have some agents to take VIP calls prior to the basic skill, but have them at the same level on the basic skill there is no more room to play with…

What is a Call Flow?

Wednesday, February 3rd, 2010

A Call Flow is a route that the call goes through.  From the phone number the caller dialed to talking to a live agent.

A Call Flow is programmed in the phone system.  Here is a simple example:

  • Caller dials 1-800-Road-Assistance
  • The 1-800 terminates in the phone system on vector x.
  • Caller hears a prompt: “Welcome to your Road Side Assistance, from service in English press 1.  Bienvenue à votre ‘assistance routière, pour le service en français, faites le 2.”
  • Caller selects 1.
  • Caller hears a prompt: “Please note your call may be recorded for quality purposes.  For Road Side assistance support, please press 1.  For service and sales services, please press 2.”
  • Caller selects 2.
  • Call sent to Skill Sales.  No agent available.
  • Caller hears announcement: “Your call is important to us, please stay on the line to maintain your calling priority.”
  • Agent available in Skill Sales, call route to agent.

May I take a second and talk about the waiting announcement?

  • “Your call is important to us, please stay on the line to maintain your calling priority.” = a true statement.
  • “Your call is important to us, you will be answered by the next available agent.” = false

I will be answered by an available agent, but the next one available?  Your phone system knows I am the next call to be answered?  If so, you are right.  If this announcement is the standard announcement, this is a flase statement.

What to consider when I create my forecast besides Historical data?

Thursday, January 21st, 2010

From previous post: “The idea is to use a model.  A model based on historical data.”

We have created a model or pattern for our call flow.  What else should we consider?  Here are some examples for our two Road Side Assistance call flow types:

Sales call flow:

  • TV promotions – What would be the impact in call volume if your Road Side Assistance company decides to advertise on TV during the Super Bowl?
    • I would suggest you meet with your Marketing department once a month, covering the next three months of promotions
  • Invoices – Is the Road Side Assistance billing always on the same day for all the customers?  Twice a month?  What would happen if instead of billing half the customers the first week of the month and half the customers the third week of the month, a decision is made to charge every customers the first week of the month?
  • What if, by mistake, customers are billed twice?  How can we plan that?  Is there a plan to change the billing application or system in three months from now?
  • Is there a new competitor with a breath taking pricing out there?
  • Did you install a new IVR (Interactive Voice response) that will allow the caller do get resolution without talking to a live agent?

Assistance support call flow:

  • At what time of the day is there more cars on the road?  Morning before work?  After work?  Nights? Week-ends?  You are right all of those would be part of the pattern based on historical data. What about Holiday week-end?
  • Is there a major weather storm planned in the next few days?  Is it more likely to have more assistance requests?

How can we forecast the number of calls we will receive if the marketing planned a promotion in the newspaper or on TV?  You need to use different sources of information.

First, currently in your Call Centre, what is the ratio of calls received / sales completed?

Then, how many sales is the marketing department planning to do based on the promotion?  At what time will the TV commercial be on the air?

Can we say we might roughly receive calls two times the number of sales planned?  (Hopefully a bit less, if the marketing was done properly, we would have a better hit rate than 2 to 1…)

Why do you need Call Centre Management?

Tuesday, January 12th, 2010

The short answer to “Why do you need Call Centre Management?” is that you need to analyze past data to forecast the future volume in your Call Centre.

To do so, you need to create a model.  Here is a simple example.

Let’s take an ice cream sandwich manufacture and a road side assistance call centre.

Ice cream sandwich manufacture:

  • Open 24/7
  • Objective: 21 000 boxes / week

Road side assistance call centre:

  • Open 24/7
  • Objective: 80/20 (We will talk about this soon!)
  • Two incoming call flows:
    • Sales
    • Assistance needed

For the ice cream sandwich manufacture, the management team would create schedules around the objective.  One of the possible solutions is to have three 8-hours shift per day with the objective of 1 000 boxes per shift.  Which would mean 3 000 boxes a day and 21 000 boxes a week.

For the Road side assistance Call Centre needs to “wait” for the calls to come in.  To a certain extent, the Call Centre do not control how busy it will be.  The idea is to use a model.  A model based on historical data.

For this example, I have created the below model for some intervals in a day.  In real life, it would have been an average day, per interval, including call volume and AHT – Average Handle Time.  (We will talk about this soon!)

As you can see, the call pattern for an average day where the morning starts slow, growing busier, slower around lunch time, then growing busier in the afternoon and finally slowing down at the end of the day.

In this case, we would would obviously try to set the lunch breaks at the middle of the day.  Your staffing model (graph curve) would need to follow the model curve.

If your are scheduling your call centre staff with a pen and paper vs a workforce management application, you would need to create many different models.

Here are some exmaples:

Average Monday, average Tuesday, average Wednesday, (…)  Those daily models would then bring you to a day of the week factor.  Through out the year, your call volume will change.  You will not receive the same call volume every month.  Your call volume and your AHT will differ monthly and weekly.

After gathering your interval data every day, you will get to a point where you have a pattern you can apply to your schedules.  Obviously, you will not receive the same number of calls you gathered from your historical data but you will have at least the model on how the day/week/month will look like.

Next post: What else should I consider besides historical data?

Luc Denis