New Creation Flow: Structure Your Topics and Improve Your Analysis

In August, Feedier introduced a new guided 3-step flow that transforms the way topics are created. The goal: make the process more intuitive, more accurate, and more reliable, so your text analysis gains clarity and relevance right from the start.

With this new flow, your teams can define a topic with context, preview related verbatims, and automatically generate refined AI instructions. No more guesswork: each topic is built on a solid foundation, aligned with your operational priorities. This update addresses a major challenge for CX teams: saving time while improving the quality of insights. It helps create stronger topic structures, reduces the risk of misclassification, and delivers a more actionable view of the Voice of the Customer.

Step 1: Defining Your Topic

The first step of the new flow is to set a clear and structured foundation for your topic. Instead of creating a topic “blindly,” you now have dedicated fields to give the AI the right context from the start:

• Name your topic (required): choose a simple and explicit name that reflects the issue or theme you want to monitor, such as “Waiting Time” or “In-store Service Quality.”
• Assign it to a topic segment (optional): segments allow you to group multiple related topics under a broader category, making it easier to track cross-functional issues without overcomplicating your structure.
• Add custom instructions (optional): provide guidance to the AI on how to match verbatims. For example, you can specify to prioritize certain keywords, ignore a particular context, or clarify the scope of the theme.

This first step is more than just filling in a form. By defining the topic with clarity and context, you ensure stronger consistency across your analysis and more accurate AI results. It aligns your topics with your operational priorities and prevents misclassification or noise in later stages of analysis.

Step 2: Review Related Verbatims

Once your topic is defined, Feedier immediately shows you a preview of the verbatims that would be matched. This real-time validation is a key improvement, as it gives you visibility and control before the topic is even created. You can then decide how to handle the proposed verbatims:

• Reject those that don’t fit: remove any irrelevant verbatims that could distort the analysis.
• Keep the others by default: if no action is taken, the verbatims remain linked to the topic.

This review step is crucial for building confidence in the system. Instead of relying on assumptions, you start with concrete examples of what the AI considers relevant. It ensures that each topic is anchored in real customer feedback, validated by your team. In practice, this step saves time for analysts and increases reliability, reducing the need for later corrections and making the analysis of customer feedback more aligned with field reality.

Step 3: Instruction Suggestions

The final step uses AI to automatically generate refined instructions based on the verbatims you accepted or rejected in the previous step. These instructions guide the system to make more accurate and consistent matches in the future. At this stage, you can:

• Accept the suggestions if they already fit your needs,
• Edit or adjust them to add your own business rules,
• Reject them if they are not relevant.

By turning your validation choices into actionable guidance, this step ensures that every new topic is built on solid rules rather than generic assumptions. The result is a more reliable topic structure, better aligned with your operational context, and a text analysis that quickly becomes more precise over time.

A new step towards more reliable text analysis

With this new guided flow, creating a topic becomes simpler, more precise, and more reliable. You can validate verbatims before linking them, let the AI generate refined instructions, and gain clarity in your text analysis from the very beginning.
Available starting with the September release, this update makes it easier to structure your topics and turn feedback into concrete actions, without unnecessary complexity.