What is Average Handle Time?

Average Handle Time is the most watched number on most contact centre dashboards, and the most mistreated. It does one job superbly: it converts messy human conversation into a unit of work that can be forecast, staffed and priced. It does another job terribly: judging whether the work was any good. Most of the damage AHT causes comes from asking it the second question when it can only answer the first.

What the number contains

AHT is the average time a contact consumes from the moment an agent takes it to the moment they are ready for the next one. It has three components:

• Talk time. The live conversation itself.
• Hold time. The customer waiting while the agent looks something up, consults a colleague or wrestles a system.
• Wrap time. The after-contact work: notes, coding the contact, raising follow-ups.

The arithmetic is simple and purely illustrative: a contact with six minutes of talk, one minute on hold and two minutes of wrap consumed nine minutes of handling. Average that across every contact in the period and you have AHT.

The decomposition matters more than the average, because the three components move for different reasons. Talk time reflects the complexity of the contact and the confidence of the agent. Hold time mostly reflects the systems and knowledge the agent has to reach for. Wrap reflects process design. An AHT rising because hold is rising is a tooling problem. An AHT rising because wrap is rising is a process problem. Treating either as an agent speed problem misses the cause entirely, and a single average hides which component moved.

A cost lens, not a quality lens

AHT earns its place on the dashboard as arithmetic. Multiply it by forecast volume and you have workload. Divide workload by what one person can carry in an interval and you have the staffing requirement. Multiply through by loaded cost and you have most of the operation’s cost base, which is why AHT sits underneath every serious read of cost-to-serve. Small movements in the average, multiplied across large volumes, are real money in either direction.

Average Handle Time prices the work. It says nothing about whether the work was done. A short call that leaves the problem unsolved is cheap on this metric and expensive in reality, because the customer comes back, angrier, to a queue that now carries the same problem twice. The quality reading has to come from somewhere else, chiefly first contact resolution, and the two numbers only mean something read together.

How the metric gets gamed

Because agents control AHT more directly than almost any other number they are measured on, turning it into a target changes behaviour immediately, and rarely in the intended direction:

• Rushed calls. Customers talked over, closed out early or pushed to self-service before the problem is understood. The call ends; the problem does not.
• Transferred problems.A transfer stops one clock and starts another. The difficult contact becomes someone else’s number, usually a more expensive someone’s.
• Mis-coded time. Wrap done while marked available, or parked in auxiliary states that fall outside the calculation. The work still happens; the metric stops seeing it.
• Cherry-picked contacts. Where agents have any influence over what they take next, the easy contacts get taken and the hard ones wait.

Every one of these has the same shape: the work is not reduced, it is moved. It reappears as repeat contacts, escalations, longer queues elsewhere and quiet damage to the customer relationship. The reported AHT falls while the true cost per resolved problem rises, which is the exact opposite of what the target was meant to achieve.

One concrete example

Clearly illustrative, with no customer implied. A centre under margin pressure sets a team-level AHT target and ties it to monthly performance conversations. The average falls steadily for six weeks and the improvement is celebrated. Then the second-line queue, which is staffed with more senior, more expensive people, starts growing, because transfers are up. Repeat contacts climb, because more first calls are ending without a resolution. Complaints follow. When someone finally measures total handling minutes per resolved problem, it is higher than before the target existed. The metric improved. The work got worse.

The decision-intelligence angle

The decision a moved AHT should trigger is almost never “make the agents faster.” It is an investigation upstream: did the contact mix shift, did a release create a new class of problem, is a new cohort still ramping, did a system slow down, did a process add wrap? Each of those points at a different owner, a different fix and a different cost, and picking between them is a real decision that deserves better than a team talking-to.

In a decision-intelligence view, AHT is evidence, not a verdict. It enters the picture with its definition stated and its components separated, it is read against resolution rather than alone, and the decisions it triggers (re-staff, fix the process, fix the product, reprice the work) are recorded with the evidence they rested on and scored when the outcome lands. A centre that manages the number gets a smaller number. A centre that manages the decisions behind the number gets a cheaper operation, which is what the number was supposed to be for.