AI-Powered Analysis

Reference

Understanding AI Confidence Scores in Text Analytics

Not every AI-classified response deserves the same level of trust. Confidence scores tell you which classifications are solid and which are worth a second look - here's what they mean and how to use them.

Tutorial

Exporting Classified Responses with Confidence Scores

Once your open-ended responses are classified, getting them out of Opionate and into a spreadsheet, client deliverable, or second review pass is one export away - including the confidence signal if you want it.

Tutorial

Reviewing and Correcting Individual AI Classifications

Editing your category set fixes the categories themselves. This is about the other kind of correction - fixing one specific response that landed in the wrong category, without touching anything else.

Reference

What Text Analytics Actually Analyzes: Your Full Response Set, Not a Sample

It's a fair question before you trust a category set built by AI: did it actually read everything, or just a sample? Here's what Text Analytics looks at when it generates categories and when it classifies.

Tutorial

Using AI-Generated Categories as a Starting Point for a Custom Taxonomy

You don't have to accept or reject an AI-generated category set as-is. Treating it as a first draft - keeping what's useful, rebuilding the rest - is often faster than either extreme.

Reference

Running Multiple Classification Passes on the Same Question

Generating categories a second time on a question doesn't erase your first attempt - it creates a new one alongside it. That's a genuinely useful safety property, and worth understanding before you regenerate.

Reference

How Many Responses Do You Need Before Category Generation Is Worth Running?

Generate categories too early, with only a handful of responses in, and you'll get a thin set that doesn't hold up as more come in. Here's how to think about timing.

Tutorial

Combining Text Analytics Results with Structured Questions in One Report

Once an open-ended question is classified, its categories behave like any other question's answer options - which means you can cross-tab open-ended themes against your structured questions in the same report.

Reference

Classifying a Response Into More Than One Category

Not every open-ended answer fits neatly into one theme. Text Analytics supports assigning a single response to more than one category when the answer genuinely touches on several - here's how it works and when to use it.

Tutorial

Auditing AI Classification Accuracy on Your Own Data

Trusting Text Analytics results for a client deliverable means being able to show your work. Here's a practical process for auditing classification accuracy on your own survey before you report the numbers.

Reference

AI vs. Manual Coding: When to Use Text Analytics and When Not To

Text Analytics isn't the right tool for every open-ended question. Here's a practical way to decide when AI-assisted coding earns its keep and when reading the responses yourself is genuinely faster.

Reference

AI Follow-up Questions: What Triggers One and What Doesn't

A follow-up doesn't fire on every open-ended answer - it fires when a condition you set is met. Here's how to configure when AI Follow-up actually kicks in, and how to keep it from firing too often or not at all.