Tuesday, May 19, 2009

Definition of a World Class Call Center? Not as bad.

There was a recent discussion on LinkedIn on what constituted world class customer service.  Was it Apple?  Zappos?  Both have fiercely loyal customers and many do report experiencing a WOW factor when they call in.
While WOW is good, I felt the discussion did not focus enough on accuracy.  In a kind of Maslow's Heirarchy of Needs, I believe a center would be better served getting the basics right (right steps to ID the customer, right diagnostic steps, right disclosures, right offer, right price, right system updates, etc) before worrying about agent warmth and empathy and any kind of WOW factor.

But go into centers and ask to see a chart of error rates on some of those dimensions listed above. They don't exist. (
http://ifyouwanttoscream.blogspot.com/2009/05/benchmark-error-rates-for-contact.html ). A call center leader for a division of a high tech company was lamenting that they could not get their thousands of outsourced agents to 1) consistently diagnose a high volume tech support call correctly, 2) they couldn't get those agents to consistently check the warranty to avoid unauthorized returns, and 3) they could not get the agents to remind the customer to remove software before returning the unit since they would never see it again, which resulted in angry letters about missing software to the Office of the CEO. Did they or their outsourcer track any of these sub-process error rates? They did not. (See Call Center Hidden Factories http://www.nationalcallcenters.org/pubs/In_Queue/vol3no7.html#Call_Center_Hidden_Factories )

Not only aren't error rates tracked, but the very process of making process changes...which happen constantly in call centers...is a complete joke (see Inside Jokes essay
http://www.nationalcallcenters.org/pubs/In_Queue/vol3no15.html ). As an example here, a client of ours went to the Philippines to listen to their outsourcer's agents take calls. In ten out of 10 calls they listened to, the agents gave the wrong price for a service that recently changed. Our client asked the management, how they communicated the change and they said what any call center management team would say: we had team meetings, we sent out emails, we did chair-drops, we monitored some phone calls and did some coaching. We have to stop kidding ourselves about our ability to change agent behavior in any kind of timely fashion (see Wag the Dog: Why are we letting agent traits control call center outputs? http://ifyouwanttoscream.blogspot.com/2009/05/wag-dog-why-are-we-letting-agent-traits.html )

I am not sure what the call center definition of World Class is, but I do know this: whether outsourced or in-house, call centers make too many errors every day (see Do Call Centers Need to Carry Malpractice Insurance
http://www.nationalcallcenters.org/pubs/In_Queue/vol2no24.html ). This is not just a few aberrant centers. The high error rates we observe are actually endemic to the call center service delivery model that most centers follow...unaided humans are not very reliable to begin with (3 Sigma at best), let alone the mostly young, entry level, low-paid employees that we staff our centers with, who (in normal economic times) are turning over at a stratoshperic rate because the jobs, in general, are stressful, who, during their short stays, don't get enough training, only occassional monitoring and even less coaching. (As an aside, coaching is the go-to method for improving call centers, but this go-to method is of questionable ROI given the high turnover (see The Futility of Call Center Coaching http://www.isixsigma.com/library/content/c080331a.asp )). With this as the typical Call Center M.O., is it any wonder we observe the quality problems that we do?

Until centers figure out how to support the intelligence and empathy of humans with the reliability of automation and blend the two together seamlessly, the high error rates in contact centers will continue and world class will only mean "not as bad."  

Friday, May 15, 2009

Wag the Dog: Why are we Letting Agent Traits Control Call Center Outputs?

I stumbled on a piece of research about how agent traits affect output measures of performance (When Conscientiousness Isn’t Enough: Emotional Exhaustion and Performance Among Call Center Customer Service Representatives).

Here is my high-level summary of the results of the study…if you use measures of conscientiousness to screen/hire, it will, in general, improve your center-wide quality scores.  However, when the agents start to get burned-out, (because of the fact that you hired agents that were more conscientious) your productivity will be even more sharply reduced. 

There are however broader implications from this study.  The paper highlights how agent conscientiousness and agent burnout affect performance.  Well, raw intelligence affects performance and degree of domain specific content knowledge affects performance and distractibility affects performance and personality affects performance and "thickness of accent" affects performance and mood affects performance and motivation affects performance and on and on and on. 

Now of course there is nothing wrong with studying employee traits to find out the ones that have the biggest effect on performance and then using that information to design selection tests to try to raise the level of performance in your centers by raising the presence of that trait.  This approach has an unassailable track record of success (see Take the Guesswork Out of Hiring) and this approach has been the bread and butter of Industrial Psychology consulting firms large (see Personnel Decisions) and small (see All About Performance) for decades.

But the bigger question is this: why are call center leaders leaving their outputs at the mercy of so many variables they can’t control?  And the industry’s attempt to deal with the challenge…to attack the endless drivers of agent variation (motivation, knowledge, conscientiousness, mood, intelligence, etc) with one-off efforts...a new selection test here, a rah-rah team meeting there, free pizza and doughnuts, occasional coaching sessions...is a fool’s errand at best. 

Agent output metrics in the call center industry will be permanently hog-tied at an embarrassingly low level until we can figure out a cost effective way to reduce the effects of between-agent variation. Selection tests help reduce this variation, but they are not enough.  Standardizing large swaths of our agents’ process using agent-assisted automation is not only the most effective and cost-efficient approach, it is the only sane solution I have seen to date.

Tuesday, May 12, 2009

Benchmark Error Rates for Contact Centers

A question was recently posted to a Linked-in User group about standards and benchmarks for call center error rates. I penned a response along these lines.


First, the direct answer to the question is there is no goal, standard, or target for acceptable error rates in call centers.  Acceptable error rates in call centers seem to be a function of what the agents are doing and what the consequences of an error are.   (For more on this, see Does the Call Center Industry Need Malpractice Insurance?)

Let's consider a situation in which we change the price for a service and we decide to check the agents' accuracy in giving the new price.  On the day after the price was changed, there is no way the agents will quote the right price 100% of the time.  What would be the acceptable error rate?  75%? 80% Would 45% be OK?  What would be acceptable two months after the price change?  90%? 95%? If the agents get this wrong it is unfortunate, but not the end of the world and most call center leaders seem willing to tolerate mediocre performance around process changes. (For more on the sloppy process changes in call centers, see Inside Jokes)

On the other hand, what would be an acceptable error rate for a disclosure on an insurance sale that is required by law or a disclosure on a credit card transaction that is also required by law where in both cases there could be federal monitors listening to calls and fining your company?  Would 95% be OK? Or would nothing short of 99.999% be acceptable?  If the latter is the case, the company would be forced to “muscle” this with lots of monitors listening to recorded calls and/or some kind of speech analytics software to check for the appropriate disclosures on every call. Five-9's quality is achievable in call centers but it comes at a steep price.

Net-net, the acceptable error rate in a contact center is what the company is willing to tolerate given the cost of errors and the cost of monitors/training/incentives to eliminate those errors.

Tracking Error Rates

The question about an acceptable error rate for call centers begs another issue:  tracking error rates.  Again, consider the example above, how many call centers would even monitor the error rates around a pricing change say a day, a week, or a month after the process change was made?  If you record every call, you can use speech analytics software to "listen" to the calls and calculate an error rate.  This is an expensive solution and not widely deployed. 

For most centers, the only way to do it is to dedicate someone to listen to 50 calls and estimate the center-wide quality rate from the sample.  Few do this.  Processes are changing all the time in call centers.  You would have to have a monitoring team almost the size of your agent population to monitor agents and track the error rates on all the process changes.  

So no one is really Tracking Error rates except on the most egregious, costly errors.   Because no one is tracking error rates, call centers commit a lot of them.

Driving Improvements

Once you determine your error rate, driving improvements is not easy.  Call monitoring is the same as inspecting in quality in manufacturing, a practice manufacturing abandoned a long time ago:  What the Call Center Industry Can Learn from Manufacturing:  Part II  The only way monitoring can drive increased compliance is if you monitor almost every call, publicly track error rates, and dismiss agents below 95%.  This is a lot of work in and of itself and it would result in a lot of expensive turnover.

A better approach is to use error-proofing and the call center equivalent of Andon lights to make it harder for agents to avoid key steps and to track quality and correct problems real time.  Desk-top consolidation and agent-assisted voice solutions are the best practices here and with these approaches, 99.999% quality is easy to achieve.  Six Sigma/Lean in Contact Centers and Agent-assisted Voice Solutions.

Monday, May 4, 2009

Mass Customization and the Transformation of the Call Center Industry

Call Centers have been around for almost four decades.  And since the first center opened, trying to find a way to raise agent output metrics while keeping costs down has been the #1 issue keeping call center leaders awake at night.  And for the most part, they have relied on coaching to get the job done.  They only problem is, it’s not working. (See The Futility of Call Center Coaching or Sorry, But Your Big Investment in Coaching and Monitoring Will Never Improve Your Output Measures.)

Leaders have wanted to leverage the Continuous Process Improvement tools such as Lean/Six Sigma, but many don’t believe they can be applied to live-agent call handling.  Why?  Because, they assume, there is no process that the agents can follow.  Why? Because, they assume, every call into a call center is unique.

Those who assume calls are unique throw up their hands and believe they can’t define a call handling process because customers call for different reasons….they ask their questions in different orders…they require different levels of explanation….they have different needs for hand-holding and small talk.  They believe that the variation in agent outputs is due to the high input variation.

Speaking of unique, Henry Ford had a unique way of dealing with input variation to his manufacturing process.  He said, “Any customer can have a car painted any colour that he wants so long as it is black.”  In other words, he stamped out input variation to keep his costs down and his factories running smoothly. 

Henry doesn’t live in our world anymore and we don’t live in his. To maximize revenue opportunities, we have to be responsive to as many of the unique needs/wants of our customers as possible.  To do so means we have to design our fulfillment operations to deliver the variation that our customers define as valuable at the lowest possible cost.  In Marketing, this capability is commonly referred to as Mass Customization.

Once car manufacturers realized they could not get away with just offering black cars, how did they respond?  They noticed that customers don’t want to vary everything.  Customers care about the color, the stereo system, sun roofs, and performance wheels.  They don’t care about bolts and axles.  When we compare one Toyota Corolla to another, much of each car is exactly the same (mass), but certain aspects of each car are unique (customization).  Also, the bolts and many sub-assemblies in a Lexus are the same ones used in a Corolla.  In other words, many parts are also shared (mass) across models (customization).

The benefits of mass customization are obvious:  you give your customers more of what they value so they keep coming back to you, while simultaneously raising the quality and lowering the costs of fulfillment.

Re-visioning the input variation in arriving calls as a mass customization opportunity opens up new solutions.  Yes there are many different types of calls, but for a given type of call, aren’t large swaths of those calls exactly the same?  Can’t we standardize those parts and just vary what the customer wants/needs us to vary?  As mentioned, car buyers care about color, not about bolts.  Similarly, customers want to talk to a live agent or get something unique about their situation addressed or they want some human empathy.  We should absolutely try to deliver that.  Do they care how a required disclosure is provided to them?  (Sotto voce:  They don’t.) 

The reason that it is so essential to start seeing the opportunities for mass customization is because to not see them, leaves us where we are today:  each incoming call is as unique as a snowflake, no process is defined for the agents to execute, and the call is riddled with sub-optimizations: 
  • Don’t automate the greeting to ensure it is branded correctly every time; hope the agents aren’t so tired and bored that they mess it up. 
  • Don’t prerecord the disclosures; hope the agents read it word-for-word without accent issues interfering with customers’ understanding. 
  • Don’t use technology to ensure the right cross-sell offer is made at the right time every time; hope your fancy variable comp plan counteracts the unyielding pressure we put on the agents to reduce their talk time. 
  • Don’t error-proof the step reminding the customer to “remove any software before returning the unit” so that it can’t be skipped; besides, the angry letters and calls from customers to the CEO about their missing software go to another department. 

Listen to a few calls a month.  Try to stem the agent variation tide with occasional coaching.  Wish the agents the best when they leave in less than a year, while you look for replacements.  Lather.  Rinse.  Repeat.

Agent-assisted voice solutions give call centers the platform to be able to deliver mass customization.  Calls are broken down into call types.  A process is 1) defined for each call type…exactly what you want the agents to do in their systems and say to customers, and 2) built using pre-recorded audio files and pre-programmed system actions.  The agent executes the process, varying it only as they need to because of something the customer says or needs, as opposed to varying everything, every time. 

The results?  Quality goes up because the agents are executing the process you gave them…no skipped steps, no accent issues, right cross-sells.  If the customer needs something unique, the agent on the call delivers it.  While service is improving, costs are going down because you are engineering the call and every agent is executing an engineered process.  In other words, you are delivering mass customization and deriving the same benefits.

The exhortation to “Think outside the box” lies somewhere between bromidic and boring, but there is real truth in it:  how we frame a problem does drive how we try to solve it.  Is every one of those millions of calls you are expecting to get this year unique? Is training the agents and hoping they get it right the best you can do?

Or are you sitting on a massive, untapped opportunity to mass customize…to give your customers more of what they really want while increasing the quality and reducing the cost of that fulfillment?

Your call.