CSAT

A CSAT form captures how a customer felt about a conversation. Each form binds to one or more channels and asks one or more questions (rating, NPS, or free text). Configure forms here; pick which form fires on close on the Close-CSAT Picker page.

Creating a form

1. Settings → CSAT Forms → + New.
2. Name (e.g. Post-resolution survey).
3. Low-rating threshold — ratings less than or equal to this trigger the supervisor low-rating alert.
4. Active — inactive forms are not delivered, but historical responses remain available.
5. Delivery channels — pick at least one of WhatsApp / Web / Email / Messenger. For each selected channel you can also tick Default form for [channel] — when multiple forms match, the one marked default for that channel wins.
6. Follow-up message — optional bilingual “thank you” sent after the customer submits.
7. Questions — add one or more questions in order. Each has:
   • Type — Rating, NPS, or Free text.
   • Key — snake_case identifier, unique within the form.
   • Prompt — bilingual (English + Arabic).
   • Scale — for Rating, pick min and max (default 1–5). For NPS, scale is locked to 0–10. Free text has no scale.
   • Required — agents/customers cannot submit without answering.

Where forms are used

• Close-CSAT Picker (Settings → Close-CSAT) — admins map each department to a default form. When an agent closes a conversation, the matching form fires.
• CraftBot CSAT node in recipes — fires the named form mid-flow.
• Agent header — agents can manually trigger a CSAT send from any conversation.

Storing the rating

Replies are parsed against the configured scale and stored on the conversation:

• The numeric rating (or text response).
• Free-form comments where the question allows.
• The timestamp of the reply.
• The agent (if any) who handled the conversation.

Responses below the form’s low-rating threshold fire a supervisor alert and feed into the low-CSAT queue.

Analytics

CSAT scores feed into the supervisor analytics dashboard:

• Rolling average per team / agent / channel.
• Distribution histogram.
• Time series.
• NPS calculation when an NPS-type question is configured (the 0–10 scale automatically computes promoters / passives / detractors).

Best practices

• Send CSAT only on resolved conversations — surveying after a Transfer-back or escalation is noisy.
• Don’t ask for CSAT on every chat in high-volume channels — sample 1 in N to avoid survey fatigue.
• Localise per language — never send an English survey to an Arabic-only customer. The bilingual prompt handles this if both fields are populated.
• Mark a default form per channel you support, so the runtime always has a deterministic choice when more than one form matches.

Channel notes

The Messenger channel is listed in the picker for forward-compatibility — outbound delivery on that channel ships when the matching channel adapter lands. WhatsApp, Web (web widget), and Email are wired today (subject to each consumer being configured).