Sentiment analysis tells you whether a comment is positive, negative, or neutral. Intent classification tells you what the commenter is trying to accomplish-and what you should do about it.
When you sort 500 YouTube comments by sentiment, you get three piles: happy, unhappy, and unclear. When you sort the same comments by intent, you get questions to answer, objections to address, feature requests to log, and support issues to route. One approach measures mood. The other reveals jobs to be done.
Why Sentiment Scores Miss the Point
Sentiment analysis works well for brand monitoring at scale. If you need to know whether a product launch generated more positive than negative buzz across ten thousand mentions, sentiment scoring does the job.
But when you're analyzing comments to inform product decisions, content strategy, or customer experience improvements, sentiment scores create more confusion than clarity.
A comment like "I've been waiting six months for dark mode" scores as negative. But the commenter isn't angry-they're invested enough to track your roadmap and remind you of a promise. That's valuable signal, not just negativity to minimize.
A comment like "This is amazing but I can't figure out how to export" scores as mixed or neutral. But the commenter has a specific, solvable problem that dozens of others probably share. Sentiment scoring buries that insight in the middle bucket.
Sentiment tells you how people feel. Intent tells you what they need and what you can do next.
The Eight Core Comment Intents
Most public comments fall into one of eight job-to-be-done categories. Each intent signals a different type of action.
Question: The commenter wants information. They're asking how something works, whether a feature exists, or what you recommend. Questions reveal knowledge gaps in your documentation, onboarding, or content.
Objection: The commenter disagrees with a claim, approach, or decision. Objections surface misalignment between your messaging and audience expectations. They're often phrased as "but what about..." or "this doesn't work when..."
Praise: The commenter is expressing appreciation or endorsement. Praise tells you what's working and which features or moments resonate most. It's also a pool of potential testimonials and case study leads.
Confusion: The commenter is lost or uncertain. They're not asking a direct question-they're signaling that something isn't clear. Confusion comments often start with "I don't understand why..." or "This seems contradictory..."
Complaint: The commenter is reporting a problem or expressing frustration with an experience. Complaints are different from objections-they're about execution, not concept. They point to bugs, friction, or unmet expectations.
Comparison: The commenter is evaluating you against an alternative. They're asking how you differ from a competitor, why they should choose you, or whether you support a feature another tool has. Comparisons reveal your positioning gaps.
Request: The commenter wants you to build, add, or change something. Requests are feature ideas, content suggestions, or asks for new formats or integrations. They're your unprompted roadmap input.
Support issue: The commenter has a technical problem that needs troubleshooting. Support issues are often phrased as "it's not working" or "I'm getting an error." They belong in your support queue, not your content backlog.
How to Classify Comments by Intent
Start with a small sample-50 to 100 comments from a single source like a YouTube video, blog post, or social thread. Read through once without tagging to get a sense of the conversation.
On the second pass, assign each comment to one of the eight intent categories. If a comment clearly serves two purposes-like a question that also includes a feature request-tag it with both intents. But most comments have one dominant job to be done.
Use a simple tagging system. A spreadsheet works. So does a Notion database, Airtable base, or any tool that lets you filter and group by tag. The format matters less than consistency.
When you're unsure which intent fits, ask: what does this person need me to do next? If the answer is "explain something," it's a question. If it's "fix something," it's a complaint or support issue. If it's "consider something," it's a request or objection.
After tagging your sample, count how many comments fall into each category. The distribution tells you what your audience needs most right now.
Intent Classification Rubric
Use this rubric to resolve edge cases and train team members on consistent tagging.
| Intent | Commenter's goal | Common phrasing | What it signals | Next action |
|---|---|---|---|---|
| Question | Get information or clarification | "How do I...", "Does this support...", "What's the difference between..." | Documentation gap, unclear messaging | Answer publicly, update docs or FAQ |
| Objection | Challenge a claim or approach | "But what about...", "This doesn't account for...", "I disagree because..." | Messaging misalignment, missing context | Acknowledge concern, explain reasoning or adjust claim |
| Praise | Express appreciation or endorsement | "This is exactly what I needed", "Finally someone gets it", "Great work on..." | What's working, potential testimonial | Thank commenter, note for case studies or social proof |
| Confusion | Signal uncertainty without asking direct question | "I don't understand why...", "This seems contradictory", "I'm not sure what you mean by..." | Unclear explanation, conflicting information | Clarify in reply, revise source content |
| Complaint | Report problem or express frustration | "This doesn't work", "I'm frustrated that...", "You promised X but delivered Y" | Execution gap, unmet expectation, possible bug | Investigate issue, route to product or support |
| Comparison | Evaluate against alternative | "How is this different from...", "Does this do what [competitor] does", "Why should I switch from..." | Positioning gap, competitive pressure | Explain differentiation, note for positioning work |
| Request | Ask for new feature or content | "Can you add...", "I'd love to see...", "Have you considered..." | Unprompted roadmap input, content idea | Log for product or content planning |
| Support issue | Need technical troubleshooting | "I'm getting an error", "It's not working when I...", "I can't access..." | Technical problem requiring diagnosis | Route to support team with context |
When a comment serves two purposes, tag both intents but identify which is primary. A comment like "I love this feature but how do I export the data?" is primarily a question with secondary praise.
Mini Case: Reclassifying 200 Tutorial Comments
A SaaS company analyzed 200 comments on a product tutorial video. Initial sentiment analysis tagged 60% as positive, 25% as negative, and 15% as neutral.
When the team reclassified the same comments by intent, the distribution looked different:
- 45 questions about features not covered in the tutorial
- 38 requests for advanced tutorials or specific use cases
- 32 praise comments highlighting favorite features
- 28 confusion signals about terminology or workflow steps
- 22 comparisons to competitor tools
- 18 complaints about missing features or bugs
- 12 objections to recommended approaches
- 5 support issues requiring troubleshooting
The sentiment analysis suggested the video performed well-60% positive is a strong score. But the intent classification revealed that nearly a quarter of commenters had unanswered questions, and another 19% wanted content the company hadn't created yet.
The team used the intent data to:
- Add timestamps and chapter markers addressing the 45 most common questions
- Create two follow-up tutorials covering the most-requested advanced topics
- Update the video description with comparison points addressing the 22 competitor mentions
- Route the 5 support issues to the customer success team
- Log the 18 feature complaints for product review
Six weeks later, new comments on the updated video showed a 40% drop in questions and a 60% increase in praise. The content became more useful not because sentiment improved, but because the team addressed what commenters actually needed.
When to Use Sentiment vs Intent
Sentiment analysis still has a place. Use it when you need to:
- Monitor brand health across thousands of mentions
- Track emotional response to a campaign or announcement at scale
- Set up automated alerts for spikes in negative sentiment
- Measure overall perception trends over time
Use intent classification when you need to:
- Turn comments into product or content decisions
- Prioritize which feedback to act on first
- Understand what your audience needs from you
- Route comments to the right team or workflow
- Identify patterns in questions, requests, or confusion
For most social research and comment analysis work, intent classification delivers more actionable insight. Sentiment scoring is a useful complement, not a replacement.
Building Your Intent Workflow
Start with one high-value source-your most-watched video, most-read post, or most-active social thread. Export the comments using platform tools or a comment export guide if you need help with the technical steps.
Tag 50-100 comments manually using the eight-category framework. This first pass trains your eye and reveals which intents dominate your audience's needs.
Once you've tagged a few hundred comments across multiple sources, you'll start to see patterns. Certain content types attract more questions. Certain topics generate more requests. Certain formats create more confusion.
Use those patterns to inform what you create next, how you structure it, and what you emphasize. Intent data turns comments from noise into a feedback loop.
If you're analyzing comments across multiple channels or need to scale beyond manual tagging, consider building a simple classification system using keyword rules or training a lightweight model on your tagged examples. But manual tagging remains the most reliable approach for small to medium datasets.
For teams new to social listening or comment research, start with the basics in our social listening starter guide before building complex classification workflows.
Sources and Further Reading
For teams working with public comments in research or compliance contexts, these resources provide additional guidance on responsible data handling and analysis practices:
- Social Media Research Ethics: Guidelines from the Association of Internet Researchers (AoIR) on ethical considerations when analyzing public social media data
- Platform Terms of Service: Review YouTube, Instagram, TikTok, and other platform policies on data use and comment analysis to ensure compliance
- Content Analysis Methods: Academic frameworks for systematic qualitative coding, available through communication and social science research journals
