Forecasting AHT (not assuming it)

Free visual lesson · about 5 minutes · short quiz at the end

ccPlanning academy · forecasting

Forecasting AHT

The block most planners assume — and why that quietly costs you.

The big idea

AHT is a forecast, not a constant.

Most plans forecast volume carefully and then drop in a single AHT — last quarter’s average, frozen for the year. But average handle time moves. Treating it as fixed builds an error into every interval.

Why it bites

AHT scales the whole requirement.

Workload is volume × AHT. A 5% AHT miss is a 5% workload miss — the same size of error as getting volume 5% wrong, but far less likely to be noticed.

It hides because nobody charts it the way they chart volume.

Driver 1

Channel and contact mix.

A day heavy on complex billing calls has a longer AHT than a day of quick balance checks. As the mix shifts — week to week, season to season — so does the blended AHT, even if no single call type changed.

Driver 2

Self-service deflection.

When the bot or app absorbs the easy contacts, what reaches an agent is the harder residue. Deflection that cuts your volume often raises your AHT — and a frozen AHT misses it entirely.

Drivers 3 & 4

Tenure and process change.

A wave of new starters lengthens AHT for months until they ramp. A new system, script, or compliance step shifts it overnight. Both are forecastable — if you’re watching AHT as a trend rather than a number.

How to do it

Forecast AHT like you forecast volume.

Trend it. Segment it by contact type and forecast the mix. Build in known step-changes (a system go-live, a hiring wave). At minimum, stop using one annual figure — refresh it on the same cadence as volume.

One more nuance

Forecast the average — but respect the tail.

Handle times are heavily right-skewed: a long tail of difficult contacts drags the mean around. You staff on the average, but that tail is what causes the intraday pain when a run of hard calls lands together.

So watch the spread, not just the mean — a stable average can hide a widening tail that makes service far less predictable.

The takeaway

Give AHT its own forecast.

It drifts with mix, deflection, tenure and process. A separate, maintained AHT forecast removes an error the size of a volume miss — one almost nobody else is watching.

Now test yourself ↓

1 / 8

Slides done? Here’s the same idea in a bit more depth — the part worth keeping.

In depth: the forecast hiding in plain sight

Workload — the thing that actually decides staffing — is volume multiplied by average handle time. Yet almost all the forecasting effort goes into the first term and almost none into the second. The volume forecast gets a model, a chart, a weekly accuracy review; AHT gets last quarter’s blended average, typed into a cell and left there for a year. Because the two terms multiply, that’s a strange place to stop: a 5% error in AHT costs you exactly as much as a 5% error in volume. The difference is that nobody is watching the AHT one, so it never gets caught.

What actually moves it

AHT isn’t a fact of nature, it’s the product of several moving parts. Contact mix shifts the blended average even when no individual call type changes — a week heavy on complex cases runs longer than a week of quick queries. Self-service deflection is the sneaky one: when a bot or app absorbs the easy contacts, the average of what’s left goes up, so the same automation that cuts your volume quietly lengthens your AHT. Tenure matters too — a hiring wave lengthens AHT for months while new starters ramp — and process or system changes can step it up overnight. Every one of those is forecastable if you’re treating AHT as a trend to be maintained rather than a constant to be inherited.

Don’t set it as a target while you’re at it

There’s a related trap worth flagging: forecasting AHT and targeting AHT are different jobs, and confusing them backfires. Most of what makes a call long — complexity, systems, process — is outside the agent’s control, so an agent-level AHT target only buys you rushed calls, skipped checks and repeat contacts. Forecast it as a planning input; manage it at the level of process and systems where the real time lives.

The principle to remember: give AHT its own forecast, refreshed on the same cadence as volume. It removes an error the size of a volume miss — one almost nobody else is even looking for.