Sentiment Analysis

Sentiment Analysis helps you understand how respondents feel about your product, service, or topic without asking them to assign a numerical rating. The AI analyzes the tone, language, and context of each answer to determine sentiment—giving you NPS-like insights without the rigidity of traditional surveys.

How Sentiment Works#

Every answer in an interview is analyzed by the AI and assigned a sentiment label with a confidence score:

• Positive: Respondent expresses satisfaction, praise, or enthusiasm
• Negative: Respondent expresses frustration, disappointment, or criticism
• Neutral: Respondent provides factual or balanced feedback without strong emotion
• Mixed: Respondent expresses both positive and negative sentiments in the same answer

The AI considers:

• Language tone: Words and phrases that signal emotion (e.g., "love", "frustrated", "amazing", "confusing")
• Context: The overall meaning of the response, not just individual keywords
• Confidence: How certain the AI is about the sentiment classification (displayed as a percentage)

Two Views: Distribution and Trend#

Distribution View#

The Distribution view shows you the overall breakdown of sentiment for the selected time period:

• Pie or Bar Chart: Visual representation of the percentage of positive, negative, neutral, and mixed responses
• Comparison to Previous Period: Shows how sentiment has changed compared to the previous period of equal length
• Color Coding: Green (positive), red (negative), gray (neutral), yellow (mixed)

Use this to answer questions like:

• What percentage of respondents are satisfied vs. frustrated?
• Has overall sentiment improved since last month?

Trend View#

The Trend view shows how sentiment changes over time:

• Time-Series Chart: Daily or weekly breakdown of sentiment across the selected date range
• Stacked or Line View: See how the proportion of each sentiment type evolves
• Anomaly Detection: Spot sudden shifts in sentiment that might indicate product issues or successful launches

Use this to answer questions like:

• Did sentiment improve after we launched the new onboarding flow?
• When did negative sentiment start increasing?

Per-Response and Per-Answer Sentiment#

Sentiment is tracked at two levels:

Overall Response Sentiment#

Each completed interview has an overall sentiment label displayed in the response list. This is calculated by aggregating the sentiment of all answers in the conversation.

Example: If a respondent gives mostly positive answers with one neutral answer, the overall sentiment is likely "positive".

Per-Answer Sentiment#

Drill into individual responses to see sentiment for each question. This helps you understand which specific topics or questions trigger positive or negative feelings.

Example: A respondent might express positive sentiment about your product's features but negative sentiment about pricing.

Keywords: Positive and Negative Themes#

The AI doesn't just assign sentiment labels—it also extracts keywords and themes that explain why respondents feel the way they do.

Positive Keywords#

Topics and themes associated with positive sentiment:

• Examples: "onboarding", "ease of use", "customer support", "intuitive design", "fast performance"
• Where they appear: In the Analytics Dashboard and in individual response summaries
• What they tell you: What respondents love about your product

Negative Keywords#

Pain points and areas of frustration:

• Examples: "slow loading", "confusing UI", "pricing", "missing features", "poor documentation"
• Where they appear: In the Analytics Dashboard and in individual response summaries
• What they tell you: What needs improvement

Using Sentiment Data#

Here's how teams use sentiment analysis to drive product decisions:

Track NPS-Like Trends Without Asking a Score#

Traditional NPS surveys force respondents to assign a 0-10 score, which can feel arbitrary. Diaform's sentiment analysis gives you similar insights by analyzing natural language—without interrupting the conversation with a rating scale.

Identify Emerging Issues Before They Escalate#

Monitor the sentiment trend chart to catch spikes in negative sentiment early. If you notice a sudden increase in negative responses, drill into the keywords and individual responses to understand what's driving it.

Example: A spike in negative sentiment with keywords like "slow loading" could indicate a performance regression that needs immediate attention.

Measure the Impact of Product Changes#

Use date filtering to compare sentiment before and after a product launch, feature release, or pricing change. Did the new onboarding flow improve sentiment? Did the price increase lead to more negative feedback?

Example: Filter responses from the 30 days before and after a feature launch to see if sentiment improved.

Prioritize Feedback Themes#

Keywords sorted by frequency help you prioritize what to work on. If "confusing UI" appears in 40% of negative responses but "missing integrations" appears in only 5%, you know where to focus first.

Next Steps#