Average handle time: forecasting it, and why AHT targets backfire
Introduction
Volume gets the attention, but average handle time is the other half of your workload — and a small error in it costs exactly as much as the same error in volume. Yet AHT is often forecast lazily, with a single blended number carried forward, and then weaponised as an agent target in a way that quietly damages the operation. This article covers what AHT is made of, how to forecast it properly, why it drifts upward over time, and why making it a target tends to backfire.
What AHT is made of
Average handle time is talk time plus hold time plus after-call work (wrap), averaged across contacts. The breakdown matters because each component has a different cause and a different fix. Long talk time usually points to contact complexity, process design, or skill. Long hold time points to slow systems or hard-to-find knowledge. Heavy after-call work points to clunky tooling or over-engineered dispositions. Treating AHT as one undifferentiated number hides all of that — you can only manage it once you’ve split it.
Forecasting it properly
The cardinal rule is the same as for volume: never blend. Forecast AHT separately by channel (voice, chat, email and back-office behave nothing alike), and ideally by call type, because a refund query and a complex complaint sit minutes apart and a shifting mix between them moves your average even when each individual type is stable. Watch the trend, not just the level — AHT has a habit of creeping up as products get more complex, processes accrete steps, and new regulation adds compliance script. And mind the distribution: handle times are heavily right-skewed (a long tail of difficult contacts), so the mean is dragged around by that tail. The average is what you staff on, but the tail is what causes the intraday pain.
Why it drifts upward
Left alone, AHT tends to rise. New regulation adds verification and disclosure. Products and propositions get more complicated. Self-service and automation skim off the simple contacts, leaving agents a harder residual mix. Each change is individually small and reasonable, which is why AHT creep is rarely forecast — and why a flat AHT assumption carried across a year of capacity planning routinely under-staffs the back half of the year. Build a modest upward trend into long-range plans unless you have a concrete reason not to.
Why AHT targets backfire
Here is the trap operations fall into again and again: they set AHT as an agent performance target. It feels intuitive — shorter calls, more calls, lower cost. But most of what makes a call long is outside the agent’s control, so the only levers an agent actually has are the harmful ones: rush the customer, skip the verification, cut the explanation, or quietly avoid the hard calls. The result is more repeat contacts (the issue wasn’t resolved), worse quality, and damaged trust — and a repeat contact is a brand-new contact in next week’s volume forecast, so you’ve made more work, not less. The very metric meant to cut cost increases it.
This connects to a broader principle: AHT is a planning input, not a behavioural target. Sometimes a longer call is the right call — one that resolves the issue first time and prevents three more. Manage AHT at the level of process, systems and skilling, where the real time lives, and leave the agent to do the job properly.
The takeaway
Split AHT into its parts and manage each at its source. Forecast it separately by channel and call type, watch the mix and the upward creep, and build trend into long-range plans. Above all, resist the urge to make it an individual target — the savings are illusory and the damage is real. Pair an honest AHT forecast with an honest volume forecast and you have the two inputs every staffing calculation depends on.
Related: the building blocks of forecasting, and how AHT feeds the Erlang C staffing calculator.