What data can be used?

As long as it's text, Gravite can work with it. The most common sources are:

• Customer reviews and verbatims — Trustpilot, Google Reviews, satisfaction surveys, NPS
• Emails and support tickets — imported from Zendesk, Intercom, or other tools
• Transcribed calls — to analyze not the quality of the agent, but the topics raised by customers
• Form responses — internal surveys, post-purchase feedback, post-delivery feedback
• Social media — mentions, comments, messages

The logic is simple: if your customers express themselves somewhere, Gravite can analyze it.

If you haven't yet connected your sources, refer to the Integrations chapter, which covers the setup of the various available connectors.

What conditions analysis quality

A few best practices to keep in mind before getting started:

• Volume: the more feedback you have over a period, the more representative and reliable the trends identified by the AI will be. A taxonomy analyzed on 50 feedbacks will produce fragile signals compared to an analysis on 5,000.
• Source consistency: analyzing Trustpilot reviews and support tickets in the same dashboard can make sense, but it's better to ensure you're comparing similar topics. Dashboard views and filters will allow for further refinement.
• Data freshness: to build your taxonomy, the AI will by default analyze your data from the last 30 days. This is a good starting point, but if your activity is seasonal or you've had a recent spike, consider adjusting the time window.

‍