​

Settings: CSAT console

CSAT | /settings/csat | Customer satisfaction survey configuration and scoring

​

1. What the CSAT console owns

• Survey template and timing. Which CSAT / NPS / CES template fires on conversation close, and the dispatch delay (1 minute to 24 hours) between resolve and send.
• Trigger scope. A smart-sentiment gate, a per-contact fatigue cooldown, and an optional condition editor that narrows dispatch by channel, assigned agent, and closing disposition.
• Scoring policy. The CSAT (1–5) and NPS (0–10) detractor bands, and whether a detractor answer auto-reopens the conversation.
• Channel eligibility. The condition editor’s channel chips gate which channels auto-dispatch fires on; all dispatch still flows through the same opt-out-checked sender every manual survey uses.

​

2. Field-by-field walkthrough

​

Master toggle — Enable auto-survey

• What it does: when on, every conversation that transitions to closed schedules a survey dispatch.
• Default: on for new tenants.
• Toggle it off to pause auto-surveys tenant-wide without losing your template or timing configuration. Manual survey sends from the API still work.

​

Survey template

• What it does: picks which template fires. The picker enumerates every row in your survey library whose type is csat, nps, or ces.
• Create templates under Insights → Surveys; the page links straight through when the list is empty.
• Edge case: if no template is selected, conversations close normally but no survey is sent — the close-arming hook reads an empty template id as “not armed.” Select a template before turning the toggle on.

​

Dispatch delay after resolve

• What it does: the minutes to wait between closed and the send. The slider is single-thumb (minutes) with a live read-out (5 min, 1h 30m, 24h).
• Clamped to 1–1440 (1 minute → 24 hours) on both save and normalisation.
• Typical: 5–30 minutes for transactional support. Longer delays reduce reply rates but feel less intrusive.

​

Smart CSAT triggering

• What it does: when on, the dispatcher skips two classes of closes before sending — conversations the sentiment pipeline labelled negative, and contacts inside your cooldown window. Off fires on every eligible close regardless of sentiment.
• Default on. Skip a negative-sentiment close on purpose: surveying a customer who just had a bad experience tends to inflame rather than collect signal.

​

Per-contact cooldown

• What it does: minimum days between surveys for the same contact, across all templates and both the CSAT and NPS tracks.
• Clamped to 7–90 days. Default 14.
• The cooldown counts any survey_responses row for the contact, so a contact who answered a survey yesterday cannot receive another today.

​

Conditional triggers

• Channels — chip toggles for the conversation channel (SMS, WhatsApp, email, voice, etc.). An empty selection means any channel.
• Assigned agents — restrict dispatch to closes resolved by specific team members. A conversation never assigned to a human (bot-only) never matches a non-empty agent list.
• Dispositions — restrict to closes stamped with specific inbox disposition labels. An empty list means any disposition.
• Exclude bot-only conversations — a per-card switch that skips conversations never picked up by a human agent.

​

Detractor rescue

• CSAT detractor threshold — on the 1–5 scale, ratings at or below this value are detractors. Default 2; clamped to 1–4.
• NPS detractor threshold — on the 0–10 scale, ratings at or below this value are detractors. Default 6 (Bain canonical: 0–6 detractor, 7–8 passive, 9–10 promoter); clamped to 0–9.
• Reopen conversation on a detractor rating — when on, a detractor answer flips the conversation back to open and tags it csat_detractor so an agent can rescue the relationship. When off, the detractor webhook event still fires; the conversation just stays closed. Archived conversations are never reopened.

​

Save

​

3. How a survey is triggered from a closed conversation

1. Close arms the dispatch. When a conversation transitions to closed, the close handler reads your console configuration. If the master toggle is on, a template is selected, and — if the condition editor is in use — the conversation matches at least one card, the handler stamps a due timestamp (now + your delay) and the template id onto the conversation.
2. The dispatch sweep fires. A scheduled sweep scans for stamped conversations whose due time has passed, re-checks smart triggering and the per-contact cooldown at send time, and dispatches through the same sender a manual survey uses — so opt-out, quiet hours, and billing behave identically to a hand-sent survey.

​

4. Scoring model

• Scale. CSAT answers use integers 1–5 (NPS sibling: 0–10). Answers outside 1–5 are rejected at intake.
• Per-response bucket. Every response is bucketed at intake as satisfied (4–5), neutral (3), or dissatisfied (1–2). The bucket travels on the response row and on the webhook payload.
• Aggregate. The answer rolls into Insights → Surveys and Insights → Sentiment rollups, and into GET /api/v1/surveys/scorecards — rows bucketed by group_by=queue or group_by=agent reporting response count, top-2-box satisfaction percent (the share of answers scoring 4–5), average score, and the NPS over NPS rows. A bucket appears only once it has at least one answered response, so unsampled sends never surface as zero-divide rows.
• Detractor handling. When a rating lands at or below your CSAT detractor threshold, it is classified detractor at intake — driving the auto-reopen playbook and the anomaly surface’s detractor-share signal.

​

5. Webhooks and API

​

Webhooks

{
  "event_type": "survey.csat.response_recorded",
  "survey_id": "8f1c0d2e-...",
  "response_id": "3c9ba5f1-...",
  "score": 2,
  "satisfaction_bucket": "dissatisfied",
  "channel": "ivr_voice",
  "agent_user_id": "7d4b1e2c-...",
  "contact_id": "c_9f3e2a1b",
  "responded_at": "2026-10-09T14:02:11Z"
}

​

API

curl "https://api.orbit.devotel.io/api/v1/surveys/scorecards?group_by=queue&survey_type=csat&from=2026-10-01T00:00:00Z" \
  -H "X-API-Key: $ORBIT_API_KEY"

​

6. Limits and edge cases

• Delay bounds. 1 minute to 24 hours (1–1440 minutes). Out-of-range values are clamped at save.
• Cooldown bounds. 7–90 days. Default 14.
• Destructive thresholds rejected. CSAT detractor threshold 0 (no detractors possible) and 5 (every rating a detractor) are rejected by the API; NPS threshold 10 is likewise rejected.
• Channel eligibility. A channel chip unselected in every condition card simply never fires on that channel — you scope out channels rather than blacklist them.
• Suppression and opt-out. Auto-dispatch flows through the same sender a manual survey uses, so an opted-out contact on a suppression list is skipped at send time — the same tenant-owned guard that gates campaign sends.
• Cooldown spans templates. The fatigue window counts survey_responses rows across CSAT and NPS, so the two tracks cannot jointly over-survey a contact.
• No template selected. Closes arm nothing and no survey fires; picking a template later does not retro-dispatch to already-closed conversations.

​

7. Troubleshooting

​

Related guides

• CSAT and NPS settings consoles — the joint CSAT/NPS read-path guide
• Auto-CSAT/NPS dispatch — the close-path dispatch loop in detail
• Surveys end to end — manual, audience-targeted distribution over the API
• Post-call surveys — IVR digit capture inside a voice call
• Settings hub — the full console map