message.comDevelopers
Developers/Tutorials/Measure CSAT and response time

Measure CSAT and response time

Pull customer satisfaction (CSAT) and response time data from the message.com reports API. Render in your own dashboard, ship to Looker / Metabase, or set SLA alerts. The numbers below assume you already have conversations and CSAT ratings flowing.

Pull first-response and resolution time

bash
curl -X GET 'https://app.message.com/api/v1/reports/tickets/response-times?range=30d' \
  -H 'Authorization: Bearer YOUR_WORKSPACE_JWT'
200 OK
{
  "range": "30d",
  "firstResponseTime": {
    "p50": 240,    // seconds
    "p90": 1800,
    "p99": 14400
  },
  "resolutionTime": {
    "p50": 7200,
    "p90": 86400,
    "p99": 259200
  },
  "byDepartment": [
    { "departmentId": "uuid", "name": "Support", "p50FRT": 180, "p90FRT": 1500 },
    { "departmentId": "uuid", "name": "Billing", "p50FRT": 420, "p90FRT": 3600 }
  ]
}

Filter by adding query params:

  • range=7d | 30d | 90d | custom (then from and to as ISO timestamps).
  • departmentId=<uuid>.
  • agentId=<uuid>.

Response time excludes off-hours when business hours are configured. p50FRT of 240 seconds means median first-reply within 4 minutes of business-hours clock.

Pull CSAT

Endpoint: GET /api/v1/reports/tickets/satisfaction?range=30d.

200 OK
{
  "range": "30d",
  "ratingCount": 412,
  "responseRate": 0.31,           // 31% of solved conversations rated
  "averageRating": 4.6,
  "byScore": { "1": 12, "2": 8, "3": 22, "4": 68, "5": 302 },
  "byAgent": [
    { "agentId": "uuid", "name": "Sam", "ratingCount": 84, "averageRating": 4.8 }
  ]
}

The default CSAT prompt is a 1-5 scale sent automatically when a conversation moves to solved. Configure under Settings → Reports → CSAT: prompt copy, scale (1-5 or thumbs up/down), delay before sending, and which channels to ask on.

Build a dashboard

Three patterns:

  • Server-side cron + database. Hit the endpoints hourly, store snapshots, render with whatever BI tool you already use.
  • Edge function + chart library. Lightweight HTML dashboard hosted on Cloudflare Pages / Vercel, fetches on view.
  • Looker / Metabase / Mode. Use the export endpoint (GET /api/v1/reports/export?range=30d&type=conversations&format=csv) on a schedule and load to your warehouse.

Set SLA alerts

Two patterns:

  • Polling. Cron job hits /reports/tickets/response-times every 15 minutes. If p90FRT > threshold, page on-call.
  • Live stream. Subscribe to the agent Socket.io namespace and watch for stale conversations (no agent message for X minutes). See Agent namespace.

Common pitfalls

  • p50 vs mean. Use percentiles. A handful of outlier tickets ruin a mean. p50 plus p90 plus p99 tells the real story.
  • Off-hours conversations. If business hours are not configured, response time counts wall-clock. Set business hours so a Saturday email does not inflate Monday numbers.
  • Low CSAT response rate. 31% is industry-typical. Lower means under-prompted or over-prompted. Tune the delay.
  • Agent attribution. If a conversation passes through three agents, "agent CSAT" can mean different things. The default attributes to the last agent before solved. Change in settings if you prefer first-agent or shared.

Next steps