Sign up to get full access to all our latest content, research, and network for everything customer contact.

Average Handle Time: The Good, the Bad and the Ugly

Add bookmark

Alex Lowenthal

Average handle time is the total amount of time spent by a call center representative in dealing with a caller’s issue, including information gathering and documentation, issue resolution/escalation and the completion of the incident processing activities even if it extends beyond the end of the call.

Call centers and service desks all over the world swear by Average Handle Time, and for good reason: There is no other real measure of efficiency.

For effectiveness, we may look at first contact resolution in the call center—however we define it; for throughput, there is call volume; and for call center quality, we may consider speed of answer, abandon rate or a call center user satisfaction survey. But for call center efficiency, average handle time is the cat’s meow. Can one even imagine a measure as simple, as straightforward and as meaningful as "2.78 minutes per call"?

We Look Upon Average Handle Time, and We See That it is Good!

On the other hand, average handle time is effectively bad since it begs so many questions: What are the contributors of average handle time? Is it the call or the after-call work that drives the number? What caused a change in average handle time? Is 2.78 minutes good or should it be less?

But what we truly need to focus on is neither its goodness, nor its evil shortcomings; our primary concern should be that average handle time is ugly. Let us survey the warts and blemishes, and discuss how to fix them.

What is a good time frame for average handle time? Too long, and it is a "mush" of no value; too short, and its variability becomes mind boggling. Consider computing average handle time by call type if you are using a menuing option for directing incoming traffic, or do so for each weekday and/or time of day (e.g. average handle time for the past six Monday mornings).

Do we fully appreciate the human errors in understanding call center measures?

Call center averages are at the mercy of outliers—one ornery or paranoid caller can ruin your call center’s average handle time.

At best, the word "average" suggests the mid-point of a bell shaped distribution in the mind of the recipient of the statistic (Everything is Gaussian, isn’t it?); at worst, average is considered a fixed value (The Gospel Truth Syndrome). Both are inappropriate for queuing models that typically use the exponential distribution for service time.

The service time distribution in most call centers is a hyper-exponential distribution.

If the issues are well understood and the responses highly scripted, then call handling time has a small variability; if, on the other hand, the issues range from dispatch (very short) to just-in-time counseling (very long) average handle time is meaningless.

Average handle time just like most measures and metrics, is a dependent variable, i.e. the result of a process or service, not the cause or controling element of that service. Do not attempt to "fix" average handle time, fix the process and see the result in the average handle time.

What is Our Purpose in Measuring Average Handle Time?

This is a good place to pay homage to Thomas Aquinas and heed his immortal warning "hominem unius libri timeo," meaning "I fear the man of a single book," or in our case, a single metric.

  • For staffing planning, the product of average call volume and average handle time should provide the necessary head-count. That is simplistic and usually inaccurate since it does not take into account hourly and daily fluctuations in call volume, or the dead time between calls, or the impact of "events" (major failures) on both call volume and handling time.
  • To demonstrate "value to the caller," we must exclude any after-call-work and measure the time from pick-up of call by the menuing system to hang-up following incident resolution; i.e. the call duration from the user’s perspective.
  • To show an increase in the value of the call center requires implementing process improvements and watching for a change in the measure of interest. For example, automated self-help enables users to perform simple tasks (e.g. password resets, account info, simple transactions, etc.). The effect of implementing this technology is a rise in average handle time as the call center addresses more complex and time consuming issues; call volume may or may not decrease—depending on whether there was a pent up demand above and beyond the volume handled by the center.
  • At the end of the day, average handle time is neither good nor bad or ugly. Those traits are only applicable to the customers of the statistic, and obviously, my learned reader, you are good.
  • Final issue: How can you explain the n dimensions and q granularities of average handle time to the executive freshly graduated from an MBA program or to the consultants he hired to "help you"? You can risk subjecting yourself to the maniacal, screeching mantra of those who prostrate themselves at the altar to Deming/Drucker/Kaplan/Grove/etc. "If you cannot measure it, you cannot manage it." Or, you can just poke your head in his office, say "Average handle time is 2.78 minutes this month!" and give him a thumbs up before walking away.

If you are in a devil-may-care mood, here are some other things you can say (although I neither recommend their use, nor accept any responsibility for the fallout) and what you can expect to hear from the exalted ones:

I really need to understand our priorities: Do you think I should drive First Contact Resolution up (with resulting increase in average handle time) or control handling time at the expense of first call resolution?

Answer: We have to work smarter or Yes

If you cannot measure it, you cannot manage it. But, just because you can measure it does not mean that you can manage it. I mean, you can measure earthquakes on the Richter scale, but you sure cannot manage the tectonic plates, can you?

Answer: Not funny!

For Geeks Only: I just read about Schr÷dinger’s cat exercise. Every time I measure something, it is like looking in the box and causing the cat’s demise. I don’t want to kill logical cats anymore.'s_cat

Answer: Are you OK?

Also for Geeks Only: You realize that, according to the Heisenberg Uncertainty Principle, if we measure Average handle time, we will not be able to ascertain exactly what the First Call Resolution is.

Answer: Huh?

First published on Call Center IQ.