Survey Analytics¶

!!! info "TL;DR" Track response rates, completion rates, and sentiment analysis powered by Claude Haiku. Export data as CSV or JSON. View real-time analytics as responses come in.

Accessing Analytics¶

Admin Dashboard → Surveys → [Your Survey] → Analytics

Overview Metrics¶

Response Rate¶

Percentage of users who completed the survey after seeing it.

Response Rate = (Completed Responses / Impressions) × 100

Example: 150 completed / 1000 impressions = 15% response rate

Benchmarks:

• Excellent: >20%
• Good: 10-20%
• Poor: <10%

Completion Rate¶

Percentage of users who finished after starting.

Completion Rate = (Completed / Started) × 100

Example: 150 completed / 200 started = 75% completion rate

Benchmarks:

• Excellent: >80%
• Good: 60-80%
• Poor: <60%

Low completion rates indicate:

• Survey is too long
• Questions are confusing or sensitive
• Technical issues (e.g., mobile rendering problems)

Median Completion Time¶

Time from start to submit (median value to exclude outliers).

Benchmarks:

• Short: <3 minutes
• Medium: 3-7 minutes
• Long: >7 minutes

Aim for <5 minutes to maximize completion rates.

Abandonment Point¶

Which question caused users to drop off?

Example report:

High abandonment at Q3 suggests the question is too sensitive, confusing, or requires too much effort.

Question-Level Analytics¶

Multiple Choice¶

Response distribution:

Visualization: Horizontal bar chart.

Rating Scale¶

Average rating: 4.2 / 5

Distribution:

5 ⭐⭐⭐⭐⭐ ████████████████████ 60 (40%)
4 ⭐⭐⭐⭐   ██████████████       45 (30%)
3 ⭐⭐⭐     ████████             20 (13%)
2 ⭐⭐       ████                 10 (7%)
1 ⭐         ██                   5 (3%)

Visualization: Histogram.

NPS (Net Promoter Score)¶

Score: +42

Segmentation:

• Promoters (9-10): 70 responses (47%)
• Passives (7-8): 55 responses (37%)
• Detractors (0-6): 25 responses (17%)

NPS = % Promoters - % Detractors
    = 47% - 17%
    = +42

Benchmarks:

• Excellent: >50
• Good: 20-50
• Poor: <20
• Negative: <0

Free Text (Long)¶

Response count: 85

Word cloud: Most common words (size = frequency)

Sentiment breakdown:

Sentiment Analysis¶

Powered by Claude Haiku for free-text responses.

How It Works¶

1. User submits free-text answer
2. Claude Haiku analyzes sentiment (positive, neutral, negative)
3. Confidence score (0-1) indicates model certainty
4. Themes are extracted (e.g., "dark mode", "performance")

Example Analysis¶

Response:

"Love the new roadmap feature! It's exactly what we needed. Performance could be better on mobile though."

Sentiment:

{
  "sentiment": "positive",
  "confidence": 0.82,
  "themes": ["roadmap", "performance", "mobile"],
  "summary": "User praises roadmap feature but notes mobile performance issues."
}

Aggregate Sentiment¶

View sentiment trends over time:

Week 1: 70% positive, 20% neutral, 10% negative
Week 2: 65% positive, 25% neutral, 10% negative
Week 3: 75% positive, 15% neutral, 10% negative

Visualization: Line chart showing sentiment % over time.

Sentiment by User Segment¶

Compare sentiment across segments:

Insight: Pro and Enterprise users are more satisfied (likely due to better features or support).

Filtering and Segmentation¶

Filter by User Properties¶

Admin Dashboard → Surveys → Analytics → Filter

Available filters:

• Plan (free, pro, enterprise)
• Country (US, CA, UK, etc.)
• Device (ios, android, web)
• Created date (after 2024-01-01)

Example: View responses from pro users on iOS:

plan = "pro" AND device = "ios"

This helps identify segment-specific feedback.

Cohort Analysis¶

Compare responses between cohorts:

Insight: Power users are most engaged and satisfied.

Exporting Data¶

CSV Export¶

Admin Dashboard → Surveys → [Your Survey] → Export → CSV

CSV columns:

• Response ID
• User ID
• Submitted At
• Q1 Answer
• Q2 Answer
• ...
• Sentiment (if applicable)
• Completion Time (seconds)

JSON Export¶

Admin Dashboard → Surveys → [Your Survey] → Export → JSON

JSON structure:

{
  "survey_id": "srv_...",
  "responses": [
    {
      "id": "resp_...",
      "user_id": "user_...",
      "submitted_at": "2024-02-10T12:00:00Z",
      "answers": {
        "q1": { "answer": "iOS" },
        "q2": { "rating": 5 },
        "q3": { "answer": "Love it!", "sentiment": "positive" }
      },
      "completion_time": 120
    }
  ]
}

API Export¶

Fetch responses via REST API:

GET /api/surveys/:surveyId/responses
Authorization: Bearer ck_live_...

See API Reference for details.

Real-Time Updates¶

Analytics update in real-time as responses come in (powered by Supabase Realtime).

Example: Open the analytics dashboard while a survey is active. Watch response counts and sentiment percentages update live without refreshing the page.

Best Practices¶

Interpreting Response Rates¶

• Low response rates (<10%) → Survey is poorly targeted, too long, or survey fatigue
• High abandonment (>40%) → Identify drop-off points, simplify questions
• Low completion rates (<60%) → Survey is too long or has confusing questions

Acting on Sentiment¶

• Positive sentiment → Identify what's working, amplify it
• Neutral sentiment → Users are indifferent, dig deeper with follow-up questions
• Negative sentiment → Prioritize fixes, respond to users directly

Segmentation Insights¶

• Compare cohorts → Identify which user segments need more attention
• Track over time → Monitor sentiment trends to measure impact of changes
• Filter by properties → Find patterns (e.g., iOS users rate higher than Android)

Next Steps¶

• Create your first survey
• Set up targeting rules
• Export response data
• Build workflows from survey responses

Tip: Use sentiment analysis to identify themes without reading every response. Focus on negative sentiment to prioritize improvements.