wave background

Reflections on managing Average Handling Time (AHT).

Change Management, Customer Satisfaction Using WFM, Forecasting, Performance and KPIs, Savings and Efficiency, Training Planning

Hussein Kamel, Senior WFM Consultant at Teleopti, looks at the best ways to work with Average Handling Time. Shorter contact durations cannot be demanded from agents, but instead the AHT must be properly examined, processes adjusted and agents assisted.

Average Handling TimeWhen creating forecasts, you are predicting volumes based on historical patterns, seasonality, special events, etc. However, there is also the other elephant in the room: Average Handling Time (AHT). Yet, though AHT merits attention, it rarely seems to get its equal share of the focus in a contact center’s operations. Interestingly, a reduction of 20% in AHT is equivalent to an equal reduction in volume on your overall staffing needs.

Most organizations are looking toward reducing their call volumes through better self-service solutions, social media, and pushing out information to customers. Yet, reducing AHT is also a strategy that centers use, and should be using even more, to reduce costs and increase efficiency. It is worth noting that we see a lot of approaches to AHT which offers plenty of opportunities to streamline contact durations. The following key points are general ideas about how to approach managing AHT reduction plans.

Don’t ask agents to consciously reduce AHT: 

Yes, it is necessary for agents to see their AHT so they’re aware of how it is trending, but if you ask your agents to drop from a four-minute call to a target of a three-minute one, they will most probably do so, but at the expense of customer experience. Of course, this does not mean that higher AHT makes happier customers (that tends to be another center legend). However, it means that if agents are thinking they must cut down their call time, they will talk faster, make mistakes, brush off customers, and overall be stressed about how fast they need to hang up and get on to the next call. They won’t be focused on customer resolution or satisfaction. So, how do you approach reducing AHT?

Analyze AHT variation: 

A first step is to create a bell curve analysis of AHT. A bell curve is a tool to measure how much variation there is from the average. AHT will always follow a bell curve pattern with differing degrees of variation. Follow this link for more on bell curves.

For example, let’s assume you have a target of three minutes, and your center is achieving an average of four minutes. You are above the target by a minute. Not good. Once you look at the variation of the data using a bell curve analysis, what do you find? Are most of your agents 30 seconds above or below the four-minute average, with the same percentage of variation from the average? Does the bell shape look tight and uniform? In this case, you see low variation in the data.

Or, are some agents hovering around three minutes, whilst others are up around the five, six or seven-minute mark, thus skewing the data for everyone else to have an average of four? Does the bell shape look loose and spread out? You then have high variation in the data.

Dealing with low variation across the group:

If the data shows you have low variation for the call durations (everyone more or less obtaining the same AHT – say, 30 seconds up or down from the target), and at the same time you are off target, it means the problem is not an issue with the agents, it is rather something at the process level. Granted there will be a few agents at the far end of the curve, but just a few doesn’t matter.

So, how do you move from four to three minutes? It is by understanding which main types of calls have the highest impact on overall AHT, and trying to re-engineer the processes of those calls to make them shorter. Are the systems slow in pulling the data for the call, does the agent need to use “hold” to do something away from the desk, ask permission, find certain information or get approval from their supervisor? Do they have to follow troubleshooting steps that could be made easier? You need to unload baggage from the process itself to make the call simpler, and once you decide on what to change, you train your agents and coach them on the new changes. Once this starts being implemented, everyone begins moving together toward the target the center is required to meet.

Dealing with high variation across the group: 

The other scenario is that you have variation in the AHT data with some agents at three minutes, and others at five, six, seven, etc… Why? For many reasons.

  • Newbies – New agents will take longer to complete calls. This is very normal. Just make sure there is a plan and step goals to get them to target, and that you are evaluating their progress on a weekly basis.
  • Different agent knowledge levels – Some agents know how to get tasks done faster than others and ask less questions. It could be that one group took the full three weeks of new agent training, while the other group received just two condensed weeks because you were losing service level and had to squeeze them through. It happens to the best of centers, but it isn’t recommended. Schedule in extra training to get the agents with lower knowledge levels and competency up to the standards you require.
  • Quality scoring methods are inconsistent among evaluators – There are different people evaluating agent performance and giving varying opinions to agents on how to handle calls, this creates a difference in agent performance. This disparity can be fixed by a solid calibration process so you have minimum discrepancy between how the evaluators think different call types need to be handled. (This will be the subject of a future blog post!)
  • Check hold time – Are some agents using hold time unnecessarily? Perhaps they are using it to take a breather on busy days. Such a phenomenon has been known to happen, and goes back to the above point around the center’s ability to catch such occurrences and give the correct guidance when warranted.

A good question now would be, what if, for these agents with high AHT variation, you reengineer and improve processes for calls, just as you would for a center that has low AHT variation? Would that work? Mostly no, since the “process” is already out of control. You need to get agents back in control, and then improve the process.

If, as a metaphor, your agents are soldiers, and supervisors are officers, you cannot ask the soldiers to march from point A to point B, unless they learnt to march together, turn, and follow instructions at the request of their officers! Otherwise, it’s going be a long slow march. In real-world terms, supervisors need to be working closely with agents to see where there are problems and figure out how to get them all up to the same level, whether that is stress management or competency development.

Final reflections

To repeat for emphasis, if you tell an agent to get their AHT down from four to three minutes by next week, most probably she/he will do it, but you are not sure how, and thus you expose your customers to the law of unintended outcomes. This is like asking a chef to make an omelet for 5 people with 2 eggs. You might get one, but you have no idea what else is in there, and it probably won’t taste very good.

Remember, you want to make customers happy as much as you want to be profitable. Happy customers, along with low AHTs, come through a refined contact process and well-managed, well-trained agents.



  1. Great article! “Spread” of AHT by individual has been of interest for some time. Danger for scheduling is if you end up scheduling all the “slower” agents (in cases where you have a wide spread of AHT achievement) all together then you can end up having SLA issues

    1. Thank you Rich, Absolutely. Then the question that might comes next is why do you have the spread ? The slower ones the newer agents, or is there something else? How can we narrow that spread over time?

  2. I agree with Rich, great article!
    I´m missing out on one part regarding more skills on agents.
    I´m of the opinion that agents seems to end up with AHT closer to the skill with the longest AHT when skills are added.
    Even if a skill with generally shorter AHT is added to the agent, he/she will end up with in call time close to the skill with the longest AHT. This still with complexity, possibility to handle multi skill and other parameters taken into account.

    Do you have any thoughts regarding multiskill to AHT?

    1. Hi Hannes,

      Thank you for your comment. . !
      Generally an agents AHT across multiple skills is a weighted average of all skills. So if an agent has an AHT of 2 minutes after answering 20 calls on skill A, and 5 minutes after answering 3 calls on skill B, the average across both would be 2:25 seconds.
      Maybe in your example they answer a lot more calls of the skill with high AHT, and less calls on the skill with lower AHT, hence the overall is skewed towards the longer calls.
      Could that be happening at your center ?
      Otherwise for planning, If we are multiskilling a group, and overall AHT for planning would as explained, we will use a weighted value.

Leave a Reply

Your email address will not be published. Required fields are marked *

Contact me I want to know more