Average Hold Time

Average hold time is the mean time a caller spends on hold after an agent has answered, across the calls that were put on hold. It is not the wait before pickup. Hold happens inside the call, once the agent goes looking for something.

OA already runs a live glossary entry on hold time, and it measures the same thing. Treat this page as the deeper cut: the maths, the two denominators, and the traps.

The confusion is nearly always about where the clock starts. Queue wait before an agent picks up belongs to a different metric — speed of answer. Hold is everything that happens after the greeting.

Key takeaways

• Hold time starts after pickup, never before it.
• Two denominators exist: all answered calls, or only the calls actually held.
• The held-calls-only figure runs far higher, so always label which one you report.
• Long holds point at knowledge, systems or authority, not lazy agents.

How it works

Average hold time divides total hold seconds by a call count. The tricky part is the count you divide by. Pick all answered calls and the number looks small. Pick only the calls that were held and it jumps.

Every contact center platform stamps a hold event the moment an agent presses the button. Plenty of calls carry two or three of them.

Add every hold segment on a single call together first, then average across calls. Averaging across segments instead quietly deflates the result, because a caller held four times looks like four short waits.

Pick one denominator, write it into the reporting definition, then stop switching. Comparing this month against last month means nothing if the maths quietly changed underneath the chart.

Both call-level figures are honest — they just answer different questions.

A dashboard that skips the label invites the same argument every Monday, because operations quotes one number and the client quotes the other.

There is no outside authority setting the bar, either. Regulators do clock one thing, and only on the way out.

The FTC’s Telemarketing Sales Rule at 16 CFR 310 treats an outbound call as abandoned when nobody connects the person to a sales representative within two seconds of their completed greeting.

That clock is for outbound dialling, not inbound hold. Inbound hold time is governed by nothing but the targets an operator writes for itself, which is exactly why the definition has to be nailed down internally.

Long hold is a symptom rather than a habit. Agents hold when the answer sits somewhere they cannot reach — a locked screen, a stale script, an approval only a supervisor can give.

That is why the metric travels so far. Hold seconds sit inside handle time, they eat the service level you promised when queues back up, and they raise the odds a caller has to come back a second time.

Repeat holds also track closely with weaker first call resolution, since an agent who cannot finish the job in one conversation usually parks the caller while hunting for someone who can.

Examples

Numbers make the denominator problem obvious, and public sources make the cost of hold obvious too. The worked cases below show how the same raw seconds produce very different headline figures and very different management responses.

A worked average. Say a team answers a thousand calls in a week and holds two hundred of them — six thousand hold seconds in total. Divide by every answered call and you report six seconds. Divide by held calls and you report thirty.

The same team, two stories. Six seconds sounds like nothing on a monthly report. Thirty seconds is what those two hundred held callers actually sat through, and it is the number that surfaces in survey verbatims and coaching sessions.

The two-second contrast. Regulators fuss over the connect moment on outbound calls, then leave inbound hold entirely to the operator. If an industry treats two seconds as material on the dial, an unlabelled thirty-second inbound hold deserves scrutiny too.

That gap matters in vendor reviews. A buyer who sees a tight two-second rule on outbound work will reasonably ask why nobody has written down an inbound hold target at all.

The staffing maths. Hold is paid time — the agent stays on the clock while the caller waits.

The US Bureau of Labor Statistics puts the median hourly wage for customer service representatives at $20.59 in May 2024.

The same source projects employment falling 5 percent from 2024 to 2034, with about 341,700 openings each year. Shrinking headcount and steady churn mean you cannot simply staff your way out of long holds.

The fix is nearly always better access to answers rather than more people waiting to deliver them. A knowledge base that returns the policy in one search beats another seat on the floor.

Public-sector service work makes the same point from a different angle. The US federal customer experience programme formalised accountability through Executive Order 14058 in 2021.

That order established High Impact Service Providers and directed 17 agencies to take 36 specific actions, while the annual guidance sits in OMB Circular A-11, Part 6, Section 280. The lesson for any operator is simple.

Hold time only improves when someone owns it by name, reports it on a fixed schedule, and has the authority to change the process behind it.

Related terms

Average hold time makes far more sense beside the metrics it borders. Each term below either measures a neighbouring slice of the call or absorbs hold seconds into a larger total, so mixing them up will distort your whole scorecard.

• Hold Time: the existing OA entry covering this same measure under its shorter name.
• Average Speed of Answer: the queue wait before an agent picks up, which hold time deliberately excludes.
• Average Handle Time: the full call total that swallows talk, hold and after-call work together.
• Call Transfer: the handover that often follows a long hold when the agent cannot resolve the issue.
• Customer Effort Score: the survey measure that captures how hard the caller had to work for an answer.

FAQ

Is average hold time the same as average speed of answer?

No. Speed of answer covers the queue wait before an agent picks up, while hold time covers the pauses after the conversation has already started.

Which denominator should I report?

Report the held-calls figure when you want to understand caller experience, and the all-answered-calls figure when you want a floor-wide average. Whichever you choose, write the denominator beside the number.

What counts as a good average hold time?

No regulator sets an inbound target, so a good figure is simply one that beats your own baseline. Set the goal against your own history, your queue mix and the promises written into client contracts.

Does a callback offer change the number?

It can, because a caller who accepts a callback leaves the call instead of sitting on hold. Check whether your platform stops the hold clock at that point or keeps counting.

How do I bring average hold time down?

Give agents faster access to answers and wider approval authority, since most long holds come from information or permission the agent simply does not have.

Compare providers, benchmarks and service standards across Outsource Accelerator before you set your own hold time target.