Call Center Dashboard Template for Voice Agent KPIs

A voice agent KPI dashboard is one page that shows whether your AI agent is answering calls, resolving them, handing off cleanly and staying within budget. Built well, it reads like the call center dashboard your team already uses, with a few rows added for things only an AI agent produces, such as latency and per-call cost.

The template below has five rows and 14 charts. Every metric in it can be charted in the Retell AI analytics dashboard, and the same layout works in any BI tool if you export the call data.

What should a voice agent KPI dashboard track?

Track outcomes first, then the things that explain them. A dashboard that leads with call volume and latency tells you the agent is busy and fast, not that callers got what they needed. These are the KPIs the template uses, with the Retell metric behind each one:

• Resolution: Retell metric (measurement): Call Successful Rate (avg). Why it is on the dashboard: Did callers get what they called for.
• Containment: Retell metric (measurement): Call Transfer Rate (avg). Why it is on the dashboard: Share of calls handed to people.
• Call volume: Retell metric (measurement): Call Counts (count). Why it is on the dashboard: The base for every rate.
• How calls ended: Retell metric (measurement): Disconnection Reason (count). Why it is on the dashboard: Separates hang-ups, transfers and errors.
• Handle time: Retell metric (measurement): Call Duration (avg, p90). Why it is on the dashboard: Typical length and the long tail.
• Caller sentiment: Retell metric (measurement): User Sentiment (count). Why it is on the dashboard: Early warning on experience.
• Responsiveness: Retell metric (measurement): End to End Latency (median, p90). Why it is on the dashboard: Pauses callers notice.
• Cost: Retell metric (measurement): Combined cost (sum, avg). Why it is on the dashboard: Spend per call and per day.
• Capacity: Retell metric (measurement): Concurrency Used (max). Why it is on the dashboard: Headroom before calls wait.

The call center dashboard template, row by row

Set the dashboard time range to the last 4 weeks and turn on previous period comparison for the headline row, so every number shows its trend.

Row 1: Headline numbers

• Calls this period (Call Counts, number chart)
• Call Successful Rate (number chart)
• Call Transfer Rate (number chart)
• Combined cost, sum (number chart)

Row 2: Outcomes

• Call Successful Rate by day, broken down by agent version (line)
• Call Counts by your call_reason field, broken down by issue_resolved (column)

Row 3: Escalations

• Call Transfer Rate by day (line)
• Call Counts by Disconnection Reason (donut)
• Call Counts by your transfer_reason field (bar)

Row 4: Experience and reliability

• Call Duration, average and p90 (line)
• End to End Latency, median and p90 (line)
• User Sentiment (donut)

Row 5: Cost and capacity

• Combined cost by day, broken down by agent (column)
• Concurrency Used, max (line)

For outbound campaigns, swap Row 3 for Call Picked Up Rate and Voicemail Rate. For cost inputs and how per-minute rates add up, see the Retell pricing page.

How do you build this dashboard in Retell?

Everything happens on the Analytics tab, as covered in the analytics dashboard documentation:

1. Open the overflow menu next to Add Chart, choose Add Dashboard, then pick Call. The call or chat scope cannot be changed later.
2. Rename the tab to something specific, such as "Support agent KPIs".
3. Select Add Chart, set the Graph Type (column, bar, line, donut or number), choose the metric under Call Source Metrics, then the measurement (count, avg, median, p90 and so on).
4. Leave the chart on Follow dashboard time range, pick the Hour, Day, Week or Month bucket, and add up to five breakdowns.
5. Save the chart and drag it into its row. A row holds up to five charts.

Filters and breakdowns set in the dashboard header apply to every chart, and you need to select Save to keep them. A workspace can hold up to 10 dashboards. If you would rather describe the dashboard than click it together, Conductor on the Analytics page can propose the charts for you to accept or reject.

Which custom fields does the template need?

Three charts depend on fields you define once in the agent editor under Post call extraction. Retell fills them from each transcript after the call ends, and post-call analysis fields can be used as chart breakdowns and filters.

• Rewrite the call_successful prompt so "successful" means the caller's task was completed.
• Add call_reason as a selector using the call reasons your team already reports.
• Add issue_resolved as a boolean: was the caller's request fully handled on this call?
• Add transfer_reason as a selector: caller asked, policy requires a person, or agent could not finish.
• Rerun analysis on recent calls after you edit the success prompt, so past results match the new definition.

What alerts should sit behind the dashboard?

A dashboard is for weekly review. Alerts are for the hour something breaks. Retell's alerting watches aggregate metrics on a schedule and sends an email or webhook when a rule triggers. The documentation's own example is a rule that fires when call success rate is below 90% over the last hour, checked every 5 minutes. Rules worth setting alongside this template:

• Call success rate drops below your normal range
• Transfer call failures above zero in an hour
• Negative sentiment rate rises compared with the last cycle
• Total call cost over a window exceeds budget
• Concurrency used approaches your limit

Review the dashboard weekly, and let alerts watch it hourly. Nobody should stare at charts waiting for a drop. Set alert rules on success rate, transfer failures and cost, and use the dashboard to find causes.

How do these KPIs map to the call center metrics you already report?

Leadership will ask for the same metrics they see for the human team, so translate rather than invent new ones:

• Average handle time: Call Duration covers the AI side. Keep human AHT separate, because it usually rises once the AI takes the simple calls. The average handle time guide shows why.
• First call resolution: issue_resolved gives you per-call resolution. Add repeat callers from your call records to get FCR, as described in the FCR benchmarks article.
• Containment: calls that ended without a transfer or take-over. The call containment guide covers how to keep it honest.

If your contact center already reports in Power BI, Tableau or Looker Studio, you can build the same rows there. Retell dashboards have no API today, so pull the underlying call records through the list calls API or the call_analyzed webhook and chart them next to your human queue data. That gives you one view across AI and human agents, which is how teams running AI customer service alongside a live floor usually report.

The quickest way to see this template with real numbers is to build it on a live agent. Create a workspace, connect an agent, and add the five rows in the Analytics tab. Start free in the Retell dashboard.