Generate a report

Computes a report over your call or contact data and returns it in data, synchronously. The shape of data depends on type, and data.type echoes it. Use GET /api/reports/types for the catalog. For raw rows instead of aggregates, use POST /api/exports.

• date_from / date_to accept an ISO 8601 datetime (offsets allowed) or a bare YYYY-MM-DD date. Since 2026-09-15 a bare date is a day in your workspace’s timezone: date_from starts at 00:00 local and date_to ends at the last instant of that day local, so both ends of the range are genuinely inclusive (date_to: "2026-08-31" used to stop at midnight UTC and leave out August 31). A value that carries a time is used as the exact instant you sent. A workspace with no timezone, or an unrecognised one, falls back to the platform zone America/Bogota. The range may not exceed 366 days and is matched against the call’s (or contact’s) created_at.
• Day buckets inside a report (cost_per_call’s time series, for example) are also your workspace’s days.
• campaign_ids filters roi_analysis and campaign_comparison. contact_funnel uses only the first id, and cost_per_call and agent_performance ignore it.
• period is optional and changes nothing: when omitted the report is identical. format accepts only json (the default); any other value, including pdf, is refused with 400 before the rest of the body is checked.

These reports count CONNECTIONS, not conversations. They do not apply the no_effective_contact rule used by /api/analytics/*: a report’s denominator is dials and connections, which answers a different and equally useful question — how much of your dialling reached a human at all. For a success rate over conversations, read successRate from GET /api/analytics/overview. The two are not meant to match.

Three fields were renamed on 2026-09-16, with no change to any figure. successfulCalls was status = 'completed' (the call connected) and successRate was connected over dialled (the share of dials that were answered); neither ever measured success. The arithmetic is unchanged — only the names now say what they count. The old keys are still sent, carrying the same values, and are kept for one release: they are removed once the dashboard, the Zapier app and the MCP tools are all on the new names. No sunset date has been set.

How each report counts:

• roi_analysis: counts only completed calls with a duration, so connectedCalls always equals totalCalls. Cost is minutes × Jelliu’s internal cost per minute. The human cost is minutes × an hourly human-agent rate ÷ 60.
• campaign_comparison / agent_performance: connectionRate is the percentage of calls in the range with status completed. Rows are ranked by it, highest first — the same ordering the deprecated successRate produced, since the two carry the same value. agent_performance credits each call to the agent of its campaign, so calls without a campaign are not counted.
• cost_per_call: one point per day with completed calls, oldest first, bucketed in your workspace’s timezone (since 2026-09-15; it was the UTC day before). Days without calls are left out.
• contact_funnel: contacts created in the range, grouped by status. The known stages (pending, called, converted, dnc) come first; any other status is appended after them.

Side effects. Read-only on your data. The request is written to the audit log as report.generate.

Idempotency. Safe to retry. Nothing is stored, and the same parameters return the same figures unless new data arrived.

Webhook events. Emits audit.log_recorded to webhooks subscribed to it. See Webhooks.

Access

• Required scope: write (it is a POST). Human users need the owner, admin or member role; viewers and billing users are refused.
• Rate limit: General API — 120 requests/min per workspace on Starter or with no active plan, 200 on Growth, 300 on Business, 600 on Enterprise. See Rate limits.
• Plan: Requires an active subscription (active, trialing or past_due); without one the request is refused with 403 BILLING_ERROR (metadata.kind: subscription_required).