Skip to main content
A phone answer-time SLA states a goal such as 75% of callers waiting for an agent are answered within 30 seconds. OpenCX records every wait for an agent and grades it against the policy that applies, so you can report compliance, outcomes, and answer times by day, hour, team, and agent.

How answer time is measured

The clock starts when a caller begins waiting for an agent: when the call enters a team queue or starts ringing agents, after any menu, greeting, or AI conversation. It stops when an agent accepts the call, not when the agent first speaks. Only time inside the policy’s office hours counts, the same as chat SLAs. A policy without office hours counts all waiting time. A caller can wait more than once in a session, for example when a call is handed back to the AI agent and transferred again. Each wait is graded on its own. Each wait ends with one outcome:

Create a phone answer-time policy

Open Settings → SLA and create or edit a policy. You need SLA write permission. Add a phone answer-time target alongside the reply and resolution targets, or create a policy with only a phone target. A policy with only a phone target grades calls alone: it can’t be applied to sessions, and workflows don’t offer it. Editing a policy only affects callers who start waiting afterwards. Waits already recorded keep the grade they received, and deleting a policy keeps its name on the waits it graded.

Choose which policy applies

OpenCX grades each wait with the first policy it finds:
  1. The team’s Phone answer-time SLA, set in Settings → Teams on the team whose queue the caller is waiting in.
  2. Otherwise, the Default answer-time SLA, set in Settings → Voice & Telephony.
Calls rung without a team use the workspace default. With neither policy, waits are still recorded with their outcome and answer time, but they are not graded and do not count toward compliance.

How agents are judged

Every ring an agent receives is recorded as answered, declined, missed (their ring ran out), or canceled (another agent answered or the caller left). Agents are judged only on their own rings:
  • Answer rate = answered ÷ (answered + declined + missed). Canceled rings are not held against anyone.
  • Time to answer runs from the agent’s ring to their accept. The queue time before their ring is never charged to them.
  • Answered within target counts the agent’s answers whose own ring-to-accept time was within the target. A wait can miss the team’s SLA while the agent who finally answered it is credited.
When a teammate’s account is deleted, their past rings still count in the totals but no longer appear in the per-agent report.

Report on answer time

Ask Companion, or have it build a mini app, for the view you need: compliance against the goal, outcomes and abandon rate, average and p90 answer time, breakdowns by day, hour of day (to find understaffed hours), team, and agent. Mini apps read two reporting tables:
  • report_phone_queue_waits: one row per finished wait, with its team, outcome, wait in seconds, who answered, the policy and target it was graded on, and the result (met, missed, excluded, or empty when no policy applied or the call failed).
  • report_phone_ring_attempts: one row per ring an agent received, with its result, ring length, and the grade of the wait it belonged to.
The same figures are available in the Phone answer-time analytics API and in the MCP tools get_phone_answer_analytics_overview, get_phone_answer_analytics_breakdown, and get_phone_answer_analytics_agents. Rates are returned as fractions from 0 to 1.

Hold time

Hold time is how long a human agent keeps a caller on hold during a call. It is separate from answer time and queue waits. Each hold is recorded from the moment the agent puts the caller on hold until they resume, the call moves to a teammate or a queue, or the call ends. A call held more than once adds its holds together, and a call transferred between teams counts toward each team for its own part. Hold time and its target apply to the team’s phone policy, chosen in the same order as answer time, and the target in effect when the hold started is the one used. Mini apps read one reporting table:
  • report_phone_call_holds: one row per held call per team, with the number of holds, the total in hold_seconds, the policy and hold_target_seconds, and over_target (1 over the target, 0 within it, empty when there is no target).
Calls never put on hold have no row, so they do not lower the average. Calls from before hold time was recorded have no row either and read as no data, not zero.

Average handle time

Average handle time (AHT) is how long human agents spend on each call they handle, from the moment they answer. Time in a menu, with the AI agent, or waiting in a queue is left out. AHT = (talk time + hold time + after-call work) ÷ calls handled
  • Talk time runs from when the agent answers until they leave the call, without their holds. An agent answers when they accept the ring, when the customer picks up their outbound call, when a warm transfer reaches them, or when they join the call as a conference guest.
  • Hold time is that agent’s own holds during the call.
  • After-call work runs from leaving the call until the agent’s wrap-up ends: when they click I’m ready, when their wrap-up timer runs out (including any extra time they add), or when their next call starts. With no wrap-up time set, it is zero.
Each agent who handles a call counts once, with their own talk, hold and after-call work, under the team they answered from. A call transferred from one agent to another counts once for each. An agent who leaves a call and comes back to it counts once, with both parts added up. Conference guests have no after-call work: they go back to taking calls as soon as they leave. Internal calls between teammates are not included. A handle counts once the agent has left the call and their after-call work has ended. Calls from before handle time was recorded have no data. The maximum handle time target comes from the team’s phone policy, chosen in the same order as answer time, and the target in effect when the agent answered is the one used. Ask Companion, or have it build a mini app, for AHT by team, agent, day, week or month. The figures are available in the Phone handle time API, its breakdown, and the list of handles (one row per stint of an agent on a call, with each part in seconds), and in the MCP tools get_phone_handle_time_overview, get_phone_handle_time_breakdown, and list_phone_call_handles. Reports & Apps → Data Export (add it to the sidebar first if it isn’t there) includes phone answer columns per session: the number of waits, the first wait in seconds, who answered first, and the worst answer-time SLA result in the session. It also includes handle time columns: the number of agent handles, total talk, hold and after-call work in seconds, and the session’s AHT. These columns are off by default; tick them under Columns.

Limitations

  • Answer time measures the accept click. It does not include time the caller spent in a menu or with the AI agent before waiting.
  • Per-call evidence for workflows stays in Get Phone Call Details. Answer-time reporting aggregates waits; it does not change those fields.