Want help applying this guide?Ask the DripTell team
A customer calls about a delivery that arrived damaged. The representative understands the problem, then puts the customer on hold to check a replacement. Two minutes later, the representative returns with a clear answer. That hold was useful. If the same call needed three holds because stock, policy, and ownership lived in different places, the number would point to a process problem, not a slow employee.
Average customer hold time is worth measuring, but it should never be used alone. Measure the time after a live conversation begins, separate it from queue waiting, and review why the hold happened. Then improve access to knowledge, authority, and specialists before asking people to work faster.
Start with the right clock
Hold time begins when a representative pauses an active voice conversation and ends when the representative returns. It is not the time before anyone answers. That earlier period is queue wait or speed to answer. It is also not after contact work completed after the customer disconnects.
Microsoft defines average conversation hold time as total hold time across handled conversations divided by the number of handled conversations. If several representatives handle one conversation, their hold time is aggregated for the conversation. Microsoft also documents hold as one component of voice handle time, alongside talk and active wrap up time, in its conversation metrics guidance.
Use one stable event definition across teams and reporting periods. Include zero hold conversations in the main average if that is how your platform defines the denominator. Do not compare that figure with a held calls only average and call the movement an improvement.
The basic calculation is simple:
Average hold time = total hold seconds across handled conversations ÷ handled conversations
Document whether transfers, outbound calls, abandoned calls, and automated voice segments are included. A small filter change can move the average even when customers experience the same service.
One average is not enough
Suppose one hundred handled calls contain ten minutes of total hold time. The average is six seconds. That sounds harmless. But the same result could mean ten people waited one minute each, or one person waited ten minutes while ninety nine people were never held. Those are different customer experiences.
- 1Fix the boundaryStart after live contact begins and stop when the representative returns.
- 2Keep the denominatorUse the same handled-conversation population across periods.
- 3Expose the distributionPair the average with exposure, percentiles, and repeat holds.
- 4Name the workRecord whether the hold served knowledge, approval, consultation, or tool delay.
- 5Repair the causeImprove access, ownership, and handoff before coaching individual speed.
Keep the official average, then add a small set of diagnostic measures:
- the share of handled conversations with any hold;
- the median hold time among held conversations;
- a high percentile such as the ninetieth percentile;
- the longest single hold event and total hold per conversation;
- repeat holds and holds before or after a transfer.
Use these measures to find cases for review, not to create a public benchmark. There is no universal good hold time. A product lookup and a fraud review do not deserve the same target.
Find the work behind each hold
A reason code should describe the work, not blame the representative. Useful groups include finding account information, checking policy, obtaining approval, consulting a specialist, preparing a transfer, completing an offline task, and waiting for a slow tool. Keep the list short enough that people can use it honestly.
Review hold time by reason, queue, customer intent, outcome, transfer count, and time of day. Microsoft’s segment based queue guidance measures cumulative hold time for each engaged inbound segment. That makes a queue or transfer view useful when the same conversation moves through more than one team.
| Signal | What it may mean | First check | Safe response |
|---|---|---|---|
| Many brief holds | Information is scattered | Search paths and permissions | Put approved answers in one reachable place |
| One long hold | A specialist or approval is unavailable | Ownership and escalation path | Offer a callback or owned follow up |
| Hold rises after transfer | Context is being rebuilt | Transfer notes and customer history | Pass the reason, evidence, and next action |
| One queue is consistently higher | The work or tools differ | Intent mix and system latency | Fix the queue process before comparing people |
| Hold falls while repeat contact rises | Calls may be ending too early | Resolution and reopen evidence | Protect complete resolution over a faster number |
The table is a starting point. Listen to or read a small, privacy appropriate sample and compare the recorded reason with what actually happened. A reason such as research may hide a missing permission, a broken knowledge path, or an approval rule no one owns.
Improve the process before coaching speed
Start with the most common avoidable reason. If representatives repeatedly place customers on hold to find the same policy, repair the knowledge source and search path. If approval is the delay, clarify who may decide and when an exception needs a supervisor. If a transfer creates another hold, make the handoff carry the customer’s question, checks already completed, and the expected next step.
Do not set a rule that merely encourages gaming. A two minute cap can produce repeated short holds, unnecessary check ins, mute instead of hold, or premature transfers. The dashboard improves while the customer keeps waiting.
Teams using a support workspace should connect the call with the customer record and owner. A shared inbox can preserve the case alongside messages, while AI Calls is an early access option for controlled voice workflows. Whatever the tool, verify that call outcomes and transcripts preserve the reason and next action. Use inbox reports to compare like with like, and test the full voice channel before treating its events as production evidence.
Set a customer rule
The operating rule matters more than a round target. Tell the customer why you need time, ask permission before placing them on hold, and return with progress or a choice. If the work will take longer than a useful live wait, offer a callback or another owned follow up. Do not make the customer call again just to protect a metric.
Review the measure weekly at first. Look for a cause that the business can remove. Hold time is useful when it reveals missing knowledge, access, or ownership. It becomes harmful when it turns a system delay into an individual speed score.
Frequently Asked Questions
What is average customer hold time?
It is total hold time across handled voice conversations divided by the number of handled conversations. Confirm the exact denominator and filters used by your reporting platform.
Is hold time the same as queue wait time?
No. Queue wait happens before a representative accepts the conversation. Hold time happens after the live conversation has begun and the representative temporarily pauses it.
What is a good customer hold time?
There is no universal number. Set expectations by intent and risk, then watch hold exposure, percentiles, repeat holds, resolution, and repeat contact. Improve the cause without rewarding rushed or incomplete service.
DripTell Editorial
Practical guidance reviewed by the DripTell product and customer workflow team.
See how DripTell checks product claims, uses primary sources and handles corrections.
Editorial and source policy