What it does
Chat widget shows structured questions before the conversation starts. Answers become context variables available for routing rules and rep conversation panels.
Key facts
- Chat channel only: not available for voice, SMS, or social messaging channels
- Max 10 questions per survey
- Max 512 characters per question text
- Answer types: Single line, Multiple lines, Option set (dropdown), User consent (checkbox)
- User consent answer type supports a link using
[link text](URL)format: incorrect format renders as raw markdown text - Context variables for each question are automatically created in workstream Advanced settings when the survey is saved
- Mandatory field validation only: no conditional logic, no question branching
When to use / skip
Use surveys on virtually every chat deployment. They're the simplest way to feed intent-based routing without an AI agent, a "What do you need help with?" dropdown drives routing rules directly. Skip only if the client wants zero friction and routes everything to a single queue.
Configuration decisions
- Which questions to ask: fewer is better; every question increases pre-chat abandonment
- Whether to use User Consent: required if there's a GDPR/privacy notice requirement before the chat starts; set to Required = Yes to enforce it
- How to name the question's Survey question name field: this becomes the context variable name and should use the exact key names for customer identification:
Name,Email,Phone,CaseNumber(seefeature-record-identification.md)
Gotchas
- Survey question names must match exactly for record identification. "CustomerEmail" instead of "Email" breaks record lookups silently.
- User Consent link format is fragile. The
[link text](URL)markdown must be exact. Test in the live widget during UAT: formatting errors show as raw markdown. - No branching or conditional logic. If the client wants different follow-up questions based on an initial answer, survey can't do it: you need an AI agent or IVR.
Consultant notes
- Pre-conversation surveys are the easiest way to get intent-based routing working on day one: a simple "What do you need help with?" dropdown with four options beats a custom AI classification model for most clients at go-live. Push for simplicity here.
- Survey question abandonment is real. Every extra question loses a percentage of customers. Challenge any client who wants more than three questions: make them justify each one. The fewer questions, the higher completion.
- The survey question name matching requirement for record identification is the gotcha that breaks the most basic implementations. Build a naming convention document as part of project setup (Name, Email, Phone, CaseNumber) and enforce it: don't leave it to individual admin judgment.
Source last updated: 2026-01-20 | Worth checking again if pre-conversation survey expands to other channels or gains conditional logic