Design CSAT survey questions and follow-up rules
Use a 1–5 star rating as the headline CSAT question, add conditional follow-ups that appear only on low ratings, and configure follow-up rules so a very low score triggers immediate action instead of sitting in a report. This article covers the questionnaire structure and the Follow-Up Rules section of a survey in Settings → CSAT Surveys.
Before you start
- Admin permissions on CSAT settings
- A survey open for editing in Settings → CSAT Surveys (new or existing)
- A clear answer to “what do we want to learn?” — the survey design depends on this
Build the questions
A common starting structure:
- 1 — Rating (1–5 stars) —
How would you rate the support you received?— The headline metric - 2 — Open-ended (conditional) —
What could we have done better?— Show only when rating ≤ 3 - 3 — Rating 1–5 (conditional) —
How easy was it to get your issue resolved?— Show only when rating ≤ 3 - 4 — Open-ended (optional) —
Anything else you'd like to share?— Always visible, not required
Conditional follow-ups can ask for either a text answer or another 1–5 rating. Use text when you want detail in the customer’s own words; use a rating when you want the follow-up to be counted and averaged alongside other rating answers.
You can also use NPS (How likely are you to recommend us?, 0–10) as the headline question instead of stars. Mixing both in one survey gets long; pick one as the primary metric and stick with it.
Configure follow-up rules
Optional but valuable. In the Follow-Up Rules section:
- Low-rating threshold — when rating is ≤ N, expand the form with a follow-up prompt
- Answer type — choose Text answer for an open-ended “what could we improve?” prompt, or Rating 1–5 when you want the customer to score a specific part of the experience
- Very-low-rating action — when rating is the lowest possible, trigger an automation rule — alert a manager, create a case, tag the conversation
csat-low
Use Text answer when the goal is diagnosis: what went wrong, what was confusing, what the customer expected. Use Rating 1–5 when you want a structured follow-up metric, such as ease of resolution, agent helpfulness, or speed. Rating follow-up answers appear in response counts and average rating calculations, and they are included in CSV exports.
The very-low-rating automation pattern is high-leverage: a single furious customer’s score becomes a same-day intervention, not a quarterly retrospective discovery.
Verify it worked
Submit a test response with a low rating on a real (non-demo) conversation: the conditional follow-up should appear after the low score, and — if you configured a very-low-rating action — the automation should fire on the lowest possible rating.