Maximum delay to answer
Definition
Maximum delay to answer: the call center hold-time cap
Maximum delay to answer is the longest a caller should wait before a live agent picks up — the ceiling, not the average. Most contact centers express it in seconds and pair it with a service level target, the classic being 80% of calls answered inside 20 seconds. Anything past the cap counts as a breach.
The metric matters because it sets the worst acceptable experience, not the typical one. Average speed of answer (ASA) can look healthy at 28 seconds even when 5% of callers wait three minutes, and those are the customers who churn, complain, or call back.
It’s a sibling of ASA, service level, and abandonment rate, but it does a different job. ASA tells you the mean. Service level tells you the percentage hitting a target. Maximum delay to answer tells you how bad the worst tail gets — which is what call-center managers actually staff against.
The cap is operational and contractual. In BPO contracts, it usually sits inside the service-level agreement (SLA) alongside abandonment and quality scores, with financial penalties if breached over a measurement window.
How it works
The call clocks in the moment the caller exits the IVR (or, in some systems, the moment the call hits the queue) and stops the moment an agent answers. If that interval crosses the cap, the call is flagged. Hold time after the answer is a separate metric and doesn’t count here, despite the legacy phrasing.
Most workforce-management platforms (Genesys, NICE CXone, Five9, Talkdesk) let you set the cap as a hard threshold and report breaches in real time. A typical reporting cadence pairs the cap with a percentile target.
Industry-typical targets sit in a narrow band. The shape below is the one most BPO RFPs use:
| Channel | Typical cap | Common SLA target |
|---|---|---|
| Inbound voice (sales) | 20 seconds | 80% answered ≤ 20s |
| Inbound voice (support) | 30 seconds | 80% answered ≤ 30s |
| Emergency / safety lines | 10 seconds | 90% answered ≤ 10s |
| Premium / VIP queues | 15 seconds | 90% answered ≤ 15s |
| Web chat | 45 seconds | 80% answered ≤ 45s |
The math behind the cap is queueing theory, specifically the Erlang C formula, which models the probability a caller has to wait at all given staffing and call volume. Wikipedia’s Call centre entry walks through how Erlang C drives capacity planning, and the broader service level page covers the percentile-based framing the industry inherited from telecom.
The lever managers pull is staffing. Add one agent to a queue of 30 callers and the worst-case wait drops faster than the average, because tail latency is more sensitive to capacity than the mean. That’s why workforce planners obsess over interval forecasts in 15- or 30-minute slices, not daily averages.
Examples
In 2024, Teleperformance reported handling more than 8 billion customer interactions across 200,000 agents. At that scale, even a 1% breach rate on a 20-second cap means 80 million slow calls a year, which is why the company publishes service-level dashboards per program rather than a corporate average.
Philippine BPOs serving US healthcare payers typically run on a 30-second cap with 80/30 service level, tightening to 20-second caps during open-enrollment surges in October and November. The seasonal staffing build (often 20–40% extra headcount for eight weeks) exists specifically to defend the cap when call volume doubles.
UK financial services regulator FCA expectations have nudged caps shorter since 2023 — after consumer-duty rules made “fair value” enforceable. Banks now commonly publish 30-second targets externally to demonstrate compliance, even though internal caps run tighter.
Emergency lines run a different game. US 911 PSAPs (Public Safety Answering Points) target 90% of calls answered in 15 seconds and 95% in 20 seconds, per NENA standard 56-005. Anything above that triggers a review.
Related terms
- Average speed of answer (ASA): the mean wait time across a window, not the cap
- Service level: the percentage of calls answered inside a target time, expressed as X/Y (e.g. 80/20)
- Abandonment rate: the share of callers who hang up before reaching an agent
- First call resolution: the share of issues solved on the first contact, no callback needed
- Average handle time: how long an agent spends on each call once connected
- Call center: the parent operation that runs all these metrics
- Workforce management: the scheduling and forecasting discipline that defends the cap
FAQ
What’s a good maximum delay to answer?
For most B2C inbound voice queues, 20–30 seconds is the working band. Sales lines run tighter (often 15–20 seconds) because every breached call is a lost lead. Support lines tolerate a longer tail.
Is maximum delay to answer the same as average speed of answer?
No. ASA is the mean wait across all calls; maximum delay to answer is the ceiling you commit not to exceed. A queue can hit a 20-second ASA while still breaching a 60-second cap on 3% of calls, so the cap catches what the average hides.
How is it measured in a BPO contract?
Usually as a service-level pair: “80% of calls answered within 20 seconds, measured in 30-minute intervals, reported monthly.” Breaches trigger SLA credits or financial penalties spelled out in the contract.
What causes breaches?
The usual suspects: under-forecasting (volume came in higher than the staffing plan), shrinkage (agents in training, sick, or in adherence breaches), and outages (system or telecom failures that drain the queue’s capacity). Most centers track each cause separately.
Does delaying the answer ever help?
Slightly, and only for routing. A 2010 Harvard Business Review study showed that customer effort, not speed alone, drives loyalty, so a 25-second wait to reach the right specialist beats a 5-second wait to reach the wrong one. The cap is a ceiling, not a target to race against.
How do AI voicebots affect the cap?
Bots either pre-qualify the caller (shifting the cap definition to the human handoff moment) or fully resolve simple intents and remove the call from the queue. Most centers now report two caps: one for the bot front-end, one for the human queue behind it.
Need a partner that hits its caps and proves it? Browse our verified BPO directory — every listing flags the service-level commitments the provider underwrites.
Independent
