The principle: analyzing sentiment at multiple levels

When feedback arrives in Gravite, the AI doesn't simply assign it an overall sentiment. It proceeds in two steps:

• An overall analysis of the feedback as a whole — what is the general sentiment expressed?
• A mention-level analysis — each topic identified in the feedback is individually extracted and independently evaluated

A single piece of feedback can thus contain multiple mentions with different sentiments: a customer can be satisfied with the response time and dissatisfied with the proposed solution, all in the same message.

It is the aggregated average of all these mentions on a given theme that constitutes the satisfaction score displayed in your dashboards — far more precise and representative than a global score calculated feedback by feedback.

What is it concretely useful for?

Analysis & Insights allows you to answer questions that traditional tools cannot handle at scale:

• What are the main contact reasons of my customers?
• On which topics is dissatisfaction the strongest?
• Which themes are improving or declining over time?
• Where should improvement efforts be prioritized?

Use cases are broad: analysis of after-sales feedback, understanding of friction points on a digital journey, tracking the perception of a new feature, detection of weak signals such as churn risk or an emerging regulatory issue.

How it works in three steps

The module is built around a simple sequence:

1. Build a taxonomy — you define with the AI the themes you want to monitor on your data
2. Attach the taxonomy to a dashboard — to visualize results in real time
3. Analyze and act — navigate through categories, track trends, understand associated verbatims

‍