Best Practices for using AI metrics and scores in Thematic

Thematic can include any score or metric from your data, such as NPS, CSAT, or AI-generated sentiment and synthetic scores. These help you see how scores trend over time and how customer feedback themes influence them.

This guide explains how to get the most accurate and actionable insights from your metrics.

Using Traditional Scores (NPS, CSAT)

Traditional scores come directly from customers, making them the most transparent and reliable when used with Thematic's Impact and Score Change features.

Themes: Fully editable and transparent in the Themes Editor.

Scores: Collected directly from customers, so always accurate.

This combination means you can trust insights like:

"Price is decreasing NPS by 2 points."

You can verify that:

• Price is an accurate theme, since it can be reviewed and edited.
• NPS is a direct customer score.
• Impact is calculated transparently.
  • Thematic's impact formula
  • Thematic's impact explanation

Using Sentiment (AI-Generated)

Thematic's sentiment analysis is automatically generated by Thematic's Internal AI system. It's consistent across all feedback sources, but only manual corrections can be made to individual comments. While sentiment is less transparent than a traditional score, it's still a valuable way to add context.

Themes: Fully editable in the Themes Editor.

Sentiment: AI-generated; consistent, but only editable per comment.

When to Use Sentiment

Sentiment works best as a layer of context alongside other metrics. For example:

• Show Detractor comments with positive sentiment.

  These customers gave a low score but weren't necessarily frustrated.

• Show 5/5 CSAT comments with negative sentiment.

  These customers rated you highly but still raised issues worth addressing.

If your stakeholders ask questions about sentiment-based insights, it helps to be upfront about what the sentiment score represents and the value it does provide to your analysis.

"Thematic's sentiment score is fully AI-generated. While it isn't as customizable or transparent as our themes, it's instantly available and consistent across all data. It adds depth that themes or customer-impact scores alone can't provide. By combining both sentiment and themes, you get a richer and more complete view of your data."

Learn more about how sentiment works in Thematic.

Using Synthetic Scores (AI-Generated, Defined by You)

Synthetic scores sit between customer scores and sentiment. Thematic's AI scores each comment, but you decide what is measured and check the results before the score is saved.

Themes: Fully editable in the Themes Editor.

Synthetic scores: AI-generated from a description and categories you define. You review a sample of comments, correct any that are wrong, and can see the accuracy before you save. You can edit the score at any time.

When to Use Synthetic Scores

• Measure something you never asked about.

  For example, cleanliness or ease of onboarding, when your survey only asks for an overall rating.

• Get a score where customers don't give one.

  For example, a Synthetic NPS on reviews or support conversations.

When sharing insights based on a synthetic score, explain how it was built and how accurate it was in review.

"This score was generated by Thematic's AI from customer comments, using a definition we wrote and checked against real feedback. It lets us measure something customers care about without asking them to rate it."

Learn how to create a synthetic score.

Tips for Sharing Insights

• Always clarify which type of metric you're using (customer-provided or AI-generated).
• Use Impact and Score Change with customer scores when you need transparent, data-driven results.
• Use Sentiment to highlight customer emotion and tone.
• Use Synthetic scores to measure specific topics customers weren't asked to rate, and share how the score was defined.
• Edit your themes and be prepared to explain the methodology for the changes
  • Thematic theme creation
  • Theme editing best practices