Themes & Sentiment
Mentions per Theme × Period × Scope with sentiment mix and rising / falling theme detection. Drills into representative quotes (PII-masked when configured). The qualitative companion to the quantitative scorecard.
Themes and experiments
Learn from customer comments and test survey wording on a stable basis.
PrimaryTheme != null
| Masked Text | Sentiment | Primary Theme | Secondary Theme | Classified By |
|---|---|---|---|---|
| MT-343 | Positive | THM-9090 | THM-9662 | AI |
| MT-975 | Neutral | THM-4152 | THM-7076 | Manual |
| MT-378 | Negative | THM-5499 | THM-9787 | Hybrid |
| MT-713 | Mixed | THM-1161 | THM-6064 | AI |
| MT-690 | Positive | THM-9896 | THM-3457 | Manual |
| MT-898 | Neutral | THM-7389 | THM-1047 | Hybrid |
Read the theme without spreading private text
Free-text feedback gives context to the score. The comment projection keeps a masked display version alongside proposed sentiment and theme labels, making it easier to group issues such as onboarding, wait time, billing or product behaviour. A comment can contribute to a primary and secondary theme without being counted as several independent customer responses.
Classification can be automated through the configured processor, but its output remains a suggestion. CX Manager can correct permitted tags with provenance. The raw text and model request follow the customer’s data policy; an alert does not copy private verbatim content into a broad notification. Anonymous comments retain no customer, ticket, agent or flow identity.
Keep a wording test stable
A wording experiment compares two published template versions within an approved campaign. The split, primary comparison, minimum sample and analysis method are agreed before delivery starts. Each invitation receives a stable variant. A delayed send, retry or reopened survey link does not move the customer to a different question set.
Response-rate comparisons use the eligible sent invitations assigned to each variant, including people who did not respond. Average-score comparisons retain their valid respondent bases. Merely reaching the minimum sample does not establish a winner. The analysis applies the prespecified decision rule and can report no significant difference or an inconclusive result. Ordinary campaign metrics remain available alongside the experiment.
Turn findings into the next revision
A rising theme can help the team choose an operational improvement, a follow-up question or a more focused audience. Keep that decision separate from editing the responses that revealed it. A corrected taxonomy or instrument is versioned so later analysis can explain what changed and when.
A concluded experiment records the comparison basis and result before a proposed wording change is promoted into a new campaign revision. Survey Author cannot bypass launch approval by selecting a successful-looking variant. CX Manager still reviews the actual instrument, audience and delivery policy. For anonymous programs, small-group and identity restrictions apply to theme and experiment outputs just as they do to scorecards. This gives your team a way to learn from feedback while keeping question design, customer privacy and the interpretation of results connected.
Modules
-
Survey design
Build CSAT, NPS and CES instruments with a clear scoring basis.
-
Campaigns and delivery
Reach eligible customers with approved surveys and controlled reminders.
-
Response capture
Keep completed answers valid, private and connected to the survey customers received.
-
Customer follow-up
Give attributable low scores an owner, due date and reviewed outcome.
-
Feedback analytics
Read CSAT, NPS and CES with the population and scoring rules behind each result.
-
Themes and experiments
Learn from customer comments and test survey wording on a stable basis.
Reports
All reportsRoles and permissions
Designs draft instruments and wording experiments for CX Manager review.
Defines analytical cohorts and interprets permitted aggregate results.
Works attributable feedback and outreach for the current customer book.
Uses attributable interaction feedback and scorecards for the current support team.
Reads privacy-governed organization trends and approved analytical snapshots.
Related processes
Survey to approved campaign
Publish the instrument and approve the exact campaign before launch.
4 stages · 1 approval
Event to accepted response
Deliver an eligible invitation and accept one valid response.
4 stages · 0 approvals
Low score to reviewed follow-up
Assign attributable customer outreach and independently review the result.
4 stages · 1 approval