Analytics
The metrics dashboard and how to read it.
The Analytics dashboard is the single place to see how your voice agents are performing across every call. It pulls together call volume, durations, answer and connection rates, call outcomes, per-agent and per-campaign breakdowns, spend trends, repeat-caller insight, and a live caller-geography globe — all on one scrolling page. This guide explains every metric and chart on the page, what each number actually measures, and how to read the data to make your agents better.
In this guide
- How to open Analytics and what plan you need
- The four headline metrics (Total Calls, Avg Duration, Total Duration, Answer Rate)
- Week-over-week comparison
- Call type, disposition and top-agent distributions
- Daily activity, peak hours and day-of-week charts
- Call outcomes, connection/abandonment and the call funnel
- Minute balance and spend trends (Minutes Per Day, Spend, Cost Per Call)
- Agent and campaign performance tables
- Repeat callers, first-time vs returning, and the caller globe
- The Deep Analytics upgrade (sentiment, keywords, transcripts and more)
- How to interpret everything to improve your agents
- Where the numbers come from, and troubleshooting
1. Opening the Analytics dashboard
Analytics lives in the Voice AI hub sidebar. From any of your portals, open the Voice section and click Analytics. The page header reads Analytics with the subtitle "Usage and performance metrics across all voice agents."
In the top-right corner there is a single circular refresh button (a circular-arrows icon). Clicking it reloads all four data sources behind the page at once, so you can pull in the most recent calls without leaving the screen.
Plan requirement
Analytics is a Growth plan and above feature. If your Voice AI plan is below Growth, the page is replaced by an upgrade screen titled "Analytics requires Growth" with an Upgrade to Growth · $79/mo button and a Compare all plans link. On the lowest tiers, you may instead see a "Feature unavailable — contact your administrator" message if your account is managed for you rather than self-serve.
Tip: There is no agent-picker or date-range selector on this page. The dashboard always shows a fixed, rolling view — the last 30 days for the activity and trend charts, the current billing period for the headline duration figures, and your most recent calls (up to roughly 500–600) for the per-call breakdowns. To narrow into a single call or agent, use the Conversations and Agents pages instead.
2. Key Metrics
The first block on the page is four large headline numbers. Above them, when available, sits a small Billing period label showing the start and end dates of your current billing cycle (for example "1 Jun – 30 Jun 2026").
| Metric | What it shows | How it is calculated |
|---|---|---|
| Total Calls | Every call handled by your agents | Sum of daily call counts over the last 30 days; falls back to the count of loaded call records if the daily breakdown is empty. Labelled "all time". |
| Avg Duration | The typical length of a call | Total talk time ÷ total calls, shown in a friendly format like 2m 14s. Labelled "per call". |
| Total Duration | All talk time added together | The total connected seconds for the current billing period, shown as e.g. 47h 51m. Labelled "this period". |
| Answer Rate | Share of calls that actually connected | Percentage of loaded calls lasting longer than 10 seconds. Labelled "calls > 10s". While the underlying call records are still loading, this shows …. |
How to read it
- Total Calls tells you raw volume. Use it to confirm your campaigns and inbound lines are actually generating traffic.
- Avg Duration is a quick health signal. Very short averages (under ~30s) often mean calls are hitting voicemail, being hung up on, or the agent is failing early. Healthy two-way conversations usually run a minute or more.
- Answer Rate is your connection signal. Anything calling under 10 seconds (rings out, instant hang-up, dead air) is treated as not answered. A low answer rate on outbound usually points to bad numbers, poor call timing, or spam-labelled caller ID.
3. Week over Week
Directly below the headline metrics is a Week over Week comparison that contrasts the current week (Sunday–Saturday) against the previous full week. Each tile shows this week's total, the previous week's total, and a percentage change with an up (▲) or down (▼) arrow.
| Tile | Meaning |
|---|---|
| Calls This Week | Number of calls so far this week vs last week. |
| Minutes This Week | Total talk minutes this week vs last week. |
| Spend This Week | Dollar spend this week vs last week (shown as $0.00). |
When the previous week had zero activity, the change indicator shows a dash (—) instead of a percentage, because a percentage change from zero is meaningless.
How to use it: Week over Week is your fastest "are things trending up or down?" check. Rising calls with flat minutes can mean shorter, lower-quality conversations. Falling spend with steady calls means your cost per call is improving.
4. Distribution Row — Call Types, Dispositions, Top Agents
Next is a three-card row that summarises your most recent calls (the loaded call sample). Each card uses thin horizontal bars with a count and percentage.
Call Types
Shows the split between inbound, outbound and web calls (and any other type the system records). The label is humanised (underscores become spaces) and each bar is sized by its share of the sample.
Dispositions
Shows the top eight call dispositions — the outcome label each call ended with (for example completed, voicemail, no_answer). Dispositions come from the per-call data; if your calls do not yet carry disposition labels, this card may show a single — group.
Top Agents
Lists your five busiest agents by call count. Each row shows the agent name, the number of calls, and the average call duration for that agent (total seconds ÷ calls).
Note: This row only appears once per-call data has loaded and there is at least one call in the sample. While loading, you'll see three shimmering placeholder cards.
5. Daily Activity — Last 30 Days
This is the main time-series chart: a smooth area/line graph of calls per day over the last 30 days. Above the chart, a summary row totals the period:
- Total calls over the 30 days
- Total duration (e.g.
132h 10m) - Total spend with the currency code (e.g.
$284.50 USD)
The chart's horizontal axis is labelled weekly (roughly every 7 days) plus the final day; the vertical axis shows call counts. Hover any point to see a tooltip with that day's date, call count, and minutes used. If there is no activity in the last 30 days, you'll see a dashed empty box reading "No call activity in the last 30 days."
6. Peak Call Hours
A 24-bar histogram showing how your calls spread across the hours of the day (12a through 11p, labelled every six hours). The tallest bars are your busiest windows. Below the bars, a Busiest hours strip lists the top three hours with their call counts (for example "2pm (188) · 11am (164) · 3pm (151)").
How to use it: For outbound campaigns, schedule dials into your proven peak windows — that's when people pick up. For inbound, make sure you have enough agent capacity (and the right routing) during the busiest hours so nobody waits.
7. Call Outcomes
A four-number block summarising the quality mix of your recent calls.
| Metric | What it shows | Threshold |
|---|---|---|
| Answer Rate | Calls that genuinely connected | Longer than 10 seconds |
| Short Calls | Quick drops or instant rejections | Under 30 seconds |
| Long Calls | Substantial, engaged conversations | Over 5 minutes |
| Total Duration | All talk time this period | — |
How to read it: A high Short Calls percentage is a red flag — it usually means voicemails, immediate hang-ups, or an opening line that isn't landing. A healthy Long Calls share shows your agent is holding real conversations. Watch these two together: you want short calls falling and long calls rising over time.
8. Connection & Abandonment
Three side-by-side pillars that classify each call by how far it got, based on call duration.
| Pillar | Definition |
|---|---|
| Connected | Calls lasting 5 seconds or more |
| Abandoned | Calls under 5 seconds (rang out, instant hang-up) |
| Voicemail | Detected by disposition where available; otherwise estimated from calls lasting 5–45 seconds. The subtitle tells you which method was used (by disposition or est. 5–45s). |
Each pillar shows a big percentage, the call count, and a thin progress bar.
9. Call Funnel
A top-down funnel showing how calls progress through stages of engagement. Each stage narrows as fewer calls reach it, with a count and percentage on the right and a short description of the drop between stages.
| Stage | Definition |
|---|---|
| Initiated | All calls attempted (100% baseline) |
| Connected | Lasted ≥ 5 seconds |
| Engaged | Lasted ≥ 30 seconds (a real conversation) |
| Completed | Lasted ≥ 2 minutes (a full call) |
| Positive outcome | Only shown when disposition data is available — calls whose disposition is not voicemail, no-answer, abandoned, busy, failed or cancelled |
If your calls don't carry disposition data, the funnel is built from duration alone, and a footnote reads "Stages based on call duration · Enable per-call sync for disposition data."
How to use it: The funnel shows you where you lose people. A big drop between Initiated and Connected is a reachability problem (bad numbers, poor timing). A big drop between Connected and Engaged is a script problem — your opening isn't holding people past the first few seconds. A drop between Engaged and Completed points to the middle of the conversation losing them.
10. Minute Balance
A summary of how many minutes you have and how many you've used. Depending on your plan, you'll see two or three columns.
| Column | Meaning |
|---|---|
| Pack Minutes | Estimated minutes remaining in your plan/pack (remaining). Shows — if not available. |
| Used This Period | Minutes consumed this billing period (minutes consumed). Pulled from the daily breakdown total, falling back to your subscription's used minutes or the period's total seconds. |
| PAYG Balance | Only shown if you have a pay-as-you-go balance above zero — your remaining wallet dollars (available). |
Tip: Cross-check Used This Period against Pack Minutes to see how much of your allowance is left and whether you're on track to run out before the period resets. For full top-up and billing controls, use the Wallet / billing pages.
11. Agent Performance
A table breaking down every agent that appears in your recent calls. Agents are sorted by call volume (busiest first), and the agent-name cell carries a small bar showing its share of the busiest agent.
| Column | What it shows |
|---|---|
| Agent | The agent (workflow) name |
| Calls | Total calls handled |
| Avg Duration | Average call length for that agent |
| Answer Rate | Share of that agent's calls lasting over 10 seconds |
| Cost | Total dollar cost of that agent's calls |
How to use it: Compare agents side by side. An agent with high call volume but a low answer rate or short average duration is the one to tune first. An agent with a high answer rate and longer durations is your model to copy.
12. Call Duration Distribution
A horizontal bar chart bucketing your calls by length, so you can see the shape of your conversations at a glance.
| Bucket | Range |
|---|---|
| < 30s | Under 30 seconds |
| 30s – 2m | 30 seconds to 2 minutes |
| 2m – 5m | 2 to 5 minutes |
| 5m – 10m | 5 to 10 minutes |
| > 10m | Over 10 minutes |
Each bar shows the count and percentage of calls in that bucket.
How to read it: A distribution heavily weighted to < 30s means many calls die early — investigate the opening line and whether you're hitting voicemail. A healthy spread into the 2m–5m and beyond buckets means your agent is sustaining conversations.
13. Campaign Performance
This table appears only when you have outbound calls. It mirrors the Agent Performance table but is filtered to outbound calls and grouped by campaign/agent name.
| Column | What it shows |
|---|---|
| Campaign | The outbound campaign/agent name |
| Calls | Outbound calls placed |
| Connected | Share of calls lasting over 10 seconds |
| Avg Duration | Average outbound call length |
| Cost | Total dollar cost of those outbound calls |
How to use it: Use the Connected column to compare list quality and timing across campaigns. A campaign with a low connection rate is wasting dials — review its contact list, caller ID reputation, and call times.
14. Calls by Day of Week
A seven-bar chart (Sun–Sat) showing which days drive the most calls. Each bar is topped with its call count.
15. Minutes Per Day & Spend Trend
Two side-by-side charts covering the last 30 days.
- Minutes Per Day — a bar chart of talk minutes per day, with the 30-day total shown above (e.g. "Total: 132h 10m last 30 days"). Hover a bar to see the date, minutes, and call count.
- Spend Trend — a smooth area chart of daily dollar spend, with the 30-day total above (e.g. "Total: $284.50 last 30 days"). Hover any point for that day's cost and minutes used.
16. Cost Per Call Trend
A smooth line chart of your average cost per call over the last 30 days (only days with at least one call are plotted). Above it sits the period average, e.g. "Avg: $0.221 per call." Hovering a point shows the day's cost per call and call count.
How to use it: Watch this for efficiency. If cost per call creeps up while average duration stays flat, something is making each call more expensive — investigate before it eats your budget. Falling cost per call with steady outcomes means you're getting more value per dollar.
17. Top Repeat Callers
A ranked list of the phone numbers that called more than once, sorted by call count (top 8). Each row shows the masked/raw number, a bar for relative volume, and the call count plus average duration.
This section only appears when caller-ID data is available for your call types (inbound calls typically have it; some web calls do not).
How to use it: Frequent repeat callers may be loyal customers, unresolved issues, or people stuck in a loop. Pull their numbers into Conversations to listen back and check whether the agent is resolving them or sending them around in circles.
18. Caller Repeat Rate (First-time vs Returning)
Two big percentages comparing how many of your unique callers are first-timers versus returners, with a split bar and a one-line summary (unique callers · returning · total calls).
- First-time — callers who appear exactly once
- Returning — callers who appear two or more times
If caller-ID isn't available for the call types in view, the section shows "Caller ID not available for these call types."
How to read it: A high Returning percentage in a support context can mean recurring problems (or great loyalty); in sales it can mean strong follow-up. Read it alongside Top Repeat Callers to understand whether returns are a good or bad sign for your use case.
19. Caller Geography Globe
An interactive globe visualising where your callers are based, derived from their phone-number country/area-code prefix. It has a left sidebar and a draggable 3D globe.
The sidebar
- A Caller Geography heading with an Origin / Location toggle.
- Origin — caller region inferred from the phone number prefix (US states by area code, plus international countries by country code).
- Location — reserved for IP-based geolocation; currently shows "IP-based caller location will appear once detected" until that data exists.
- A ranked list (top 10) of regions with a count and bar each. Numbered
01,02, … Below the list, web/no-caller-ID calls are noted (e.g. "+ 132 web / no caller ID"). - A footnote describing the basis ("Based on phone number prefix" or "Based on IP geolocation").
The globe
- A monochrome orthographic globe that auto-rotates slowly, with brighter shading on regions that have more calls and white dots marking active regions (dot size scales with volume).
- Drag to rotate, scroll (while hovering) or use the + / − buttons to zoom. A hint at the bottom reads "Scroll to zoom · Drag to rotate."
How to use it: Geography tells you where demand is concentrated. For outbound, it confirms you're dialling the right area codes; for inbound, it can reveal markets worth a local number or region-specific messaging.
20. Unlock Deep Analytics (Growth plan and above)
At the bottom of the page is a promotional panel headed "Deeper intelligence, on every call" with a Growth Plan & Above badge and a View plans button (linking to the plans page). It previews six AI-powered analytics features that go beyond the standard dashboard. Each card carries a lock icon — these are upgrade previews, not yet-active controls on this page.
| Feature | What it offers |
|---|---|
| Sentiment Analysis | Emotion and tone detection per call |
| Keyword & Topic Spotting | Surfaces recurring themes automatically |
| Full Transcript Search | Search across every conversation |
| Custom Report Builder | Build and schedule your own reports |
| CRM Deal Tracking | Auto-link calls to contacts and deals |
| Agent Performance Scoring | AI rubric-based quality scoring |
21. How to interpret the data to improve your agents
The dashboard is most useful when you read the panels together. Here are the common diagnostic patterns.
Reachability problems (calls aren't connecting)
- Symptoms: Low Answer Rate, high Abandoned in Connection & Abandonment, a big drop at the Connected stage of the funnel, lots of < 30s calls in the duration distribution.
- What to do: For outbound, check your contact list quality and caller-ID reputation, and move dials into your Peak Call Hours / strongest Day of Week. For inbound, ensure routing and capacity match your busiest windows.
Engagement problems (calls connect but die quickly)
- Symptoms: Decent Answer Rate but a steep drop between Connected and Engaged in the funnel; short Avg Duration; heavy 30s–2m bucket.
- What to do: Rework the agent's opening line and first questions. Listen to a handful of short calls in Conversations to hear exactly where people drop.
Outcome problems (calls run long but don't convert)
- Symptoms: Good Engaged/Completed numbers but a low Positive outcome stage, or dispositions skewed toward voicemail/no-answer.
- What to do: Refine the agent's goal handling and call-to-action. Compare your best- and worst-performing agents in Agent Performance and bring the laggards in line.
Cost problems
- Symptoms: Rising Spend Trend or Cost Per Call without a matching rise in good outcomes.
- What to do: Trim overly long calls, fix loops (check Top Repeat Callers), and confirm your minute usage in Minute Balance is sustainable for the period.
A simple weekly routine
- Glance at Week over Week — are calls, minutes and spend moving the right way?
- Check Answer Rate and the Call Funnel for where you're losing people.
- Scan Agent Performance and Campaign Performance for outliers.
- Confirm Minute Balance is on track for the billing period.
- Open Conversations to listen to a few calls behind any surprising numbers.
22. Where the numbers come from
Different panels draw on different data sources, which is why some appear only when certain data exists.
| Source | Powers |
|---|---|
| Daily usage breakdown (last 30 days) | Total Calls, Daily Activity chart, Week over Week, Minutes Per Day, Spend Trend, Cost Per Call, Used This Period |
| Current billing period usage | Total Duration, the Billing period label |
| Per-call records (your most recent calls, up to ~200–600) | Call Types, Dispositions, Top Agents, Answer Rate, Call Outcomes, Peak Call Hours, Day of Week, Agent & Campaign Performance, Duration Distribution, Funnel dispositions |
| Call sessions / wallet history | Connection & Abandonment, Funnel duration stages, Top Repeat Callers, Caller Repeat Rate, Caller Globe |
A few practical consequences:
- The dashboard prefers the most reliable available source and falls back gracefully. For example, if detailed per-call dispositions aren't synced, the funnel and connection panels are built from call duration instead, and you'll see a note saying so.
- Caller ID is needed for the geography globe, repeat-caller and first-time/returning panels. Call types without a caller number (some web calls) are grouped as web / no caller ID.
- Voicemail is detected from disposition labels when present; otherwise it's an estimate based on call length (5–45 seconds), so treat that figure as approximate.
23. Frequently asked questions
Can I filter by a single agent or a custom date range? Not on this page. The dashboard shows fixed rolling windows (last 30 days for trends, the current billing period for durations, and your most recent calls for breakdowns). For a specific agent or call, use Agents and Conversations.
Why is the Campaign Performance table missing? It only appears when you have outbound calls. If you only run inbound or web agents, you won't see it.
Why does a percentage show a dash (—) in Week over Week? Because the previous week had zero activity, so a percentage change can't be calculated.
Why is my Answer Rate showing "…"? The per-call data is still loading. Give it a moment or press the refresh button.
The page says "No call activity in the last 30 days." Why? Your agents haven't taken any calls in that window, or the calls haven't synced yet. Make a test call and refresh.
What does "Enable per-call sync for disposition data" mean? Your calls are currently measured by duration only. Once detailed per-call outcomes (dispositions) are synced, the funnel and outcome panels become more precise and the Positive outcome funnel stage appears.
How do I get sentiment, keyword and transcript analytics? Those are part of Deep Analytics, available on the Growth plan and above. Use the View plans button at the bottom of the page to upgrade.
24. Quick reference
| Panel | Window | Appears when |
|---|---|---|
| Key Metrics | Billing period + last 30 days | Always |
| Week over Week | This vs last week | Breakdown data exists |
| Distribution Row | Recent call sample | Per-call data loaded |
| Daily Activity | Last 30 days | Always (empty state if no calls) |
| Peak Call Hours | Recent call sample | Per-call data loaded |
| Call Outcomes | Recent call sample | Per-call data loaded |
| Connection & Abandonment | Recent calls | Duration data available |
| Call Funnel | Recent calls | Duration data available |
| Minute Balance | Billing period | Always |
| Agent Performance | Recent call sample | Per-call data loaded |
| Call Duration Distribution | Recent call sample | Per-call data loaded |
| Campaign Performance | Recent outbound calls | Outbound calls exist |
| Calls by Day of Week | Recent call sample | Per-call data loaded |
| Minutes Per Day / Spend Trend | Last 30 days | Breakdown data exists |
| Cost Per Call | Last 30 days | Days with calls exist |
| Top Repeat Callers | Recent calls | Caller ID available |
| Caller Repeat Rate | Recent calls | Caller ID available |
| Caller Globe | Recent calls | Always (empty state if no caller ID) |
| Deep Analytics upsell | — | Always |
The full metric catalogue (40+ metrics)
Analytics and the Client CRM track 40+ metrics across six categories, all updating in real time. You can export any view to CSV or schedule a report to your inbox.
| Category | Metrics |
|---|---|
| Calls | Total volume · average duration · outcome distribution · sentiment score · voicemail rate · first-call resolution |
| Leads | Lead score (1–10) · stage velocity · source attribution · qualification rate · days in pipeline · conversion rate |
| Revenue | Revenue per agent · revenue per campaign · cost per qualified lead · average deal value · MRR · won vs lost |
| Team | Calls handled · average handle time · close rate · calls-to-close · agent utilization · target vs actual |
| Campaigns | Connect rate · booking rate · campaign ROI · attempts vs completions · A/B variant performance · DNC hit rate |
| Contacts | New vs returning · total contacts · tag distribution · segments · recall-campaign performance |
Reports can be scheduled to land in your inbox automatically, and everything is exportable — there's no data lock-in.