Quick answers to common questions about Text Analytics - Opionate's tool for turning open-ended survey responses into structured, reportable categories. For a fuller walkthrough, see our Introduction to Text Analytics.
Building a Classification¶
Do I have to use AI to categorize my open-ended responses?¶
No - you can define your own categories manually, with your own names and definitions, without any AI involvement. This works at any data volume and doesn't require AI-suggested categories to get started.
How does AI-suggested categorization work?¶
You describe what you're looking for - the angle, roughly how many categories you expect, anything you already suspect might be a theme - and AI proposes a starting set of categories based on your actual responses. Our guide on writing a good classification goal covers how to phrase this well.
Is there a minimum number of responses needed for AI to suggest categories?¶
Yes, practically speaking - AI-suggested categorization needs enough textual signal to find genuine, recurring patterns. Below roughly 10-15 substantive responses, it's more reliable to define your own categories manually. Our guide on how much data is enough covers practical thresholds in more depth.
Can I edit the proposed categories before classifying responses?¶
Yes - you get a chance to review the proposed category list, rename anything, tighten a definition, add a missing category, or remove one that isn't useful, before any response actually gets sorted.
Running a Classification¶
Can a single response belong to more than one category?¶
Yes, if you set the classification up to allow it - a multi-label classification lets one response carry more than one category, which is more accurate for responses that genuinely touch multiple themes, at the cost of category percentages no longer summing to a clean 100%. Our multi-label vs. single-label guide covers how to choose and how to read the results correctly.
What does the confidence score on a classified response mean?¶
It reflects how certain the classification was about that specific assignment - a high score means the response matched the category clearly; a lower score means it sat closer to a boundary between two possible categories. It's not a guarantee of correctness, but it's the most useful signal for deciding where to spend limited review time. See our guide on when to trust AI categorization.
Can I build a sentiment classification (positive/neutral/negative)?¶
Yes - sentiment isn't a separate, dedicated tool, it's a classification you define the same way as any other, with categories for positive, neutral, and negative. Our guide to building a sentiment classification covers how to write definitions that hold up.
Reviewing and Correcting¶
Can I manually fix a response that got sorted into the wrong category?¶
Yes - every classified response can be manually reassigned to a different category, both to fix individual mistakes and to build confidence in the categories before reporting on them.
How do I check whether a classification is trustworthy before I report on it?¶
Start with the lowest-confidence responses in each category and read a sample directly - even reviewing the bottom 5-10% by confidence catches a disproportionate share of genuine issues. Our guides on when to trust AI categorization and catching AI miscategorization cover a fuller review process.
Can I drill deeper into one category without redoing the whole classification?¶
Yes - you can build a second classification scoped to just the responses inside one category from an earlier pass, without touching or re-sorting anything outside it. Our second-pass classification guide covers when this is worth doing.
Can I try a different category scheme without losing my first attempt?¶
Yes - duplicate a classification to test a meaningfully different set of categories on the same responses, keeping your original attempt intact alongside the new one.
Getting Your Data Out¶
Can I export classified responses?¶
Yes - classified responses export to CSV, with each response's assigned category included, ready for further analysis or reporting elsewhere.
Can classified categories appear on a dashboard alongside my other charts?¶
Yes - classified categories can be added directly as a chart widget on the same dashboard as your quantitative data, and crossed against the same segments your other widgets use. See our guide on bringing open-ended themes into a quantitative dashboard.
Credits¶
Does building a classification cost credits?¶
Defining and editing categories yourself is free. The AI-assisted steps specifically - proposing categories from your data, and running a classification pass - draw from your credit balance. See our guide on what actually consumes your credits for the full breakdown.
Does re-running a classification after editing the categories cost credits again?¶
Rerunning the AI-assisted classification step after a category revision does draw from your balance again, the same as the original run - manual edits to individual response assignments, by contrast, are always free.