Surveys

Build Smart Surveys with AI: A Conversational Approach Using MCP

Design and deploy intelligent surveys by simply describing what you want to learn. Let Claude handle the question design, branching logic, and survey structure.

Traditional survey builders require clicking through interfaces, configuring question types, and manually setting up branching logic. With AppGram MCP and Claude, you describe your research goals and let AI design the optimal survey structure.

Benefits of the conversational approach:

• Faster creation — Describe a survey in plain English, get it built in seconds
• Better questions — Claude suggests question wording based on best practices
• Smart branching — Easily create conditional flows based on answers
• Quick iteration — Modify surveys through conversation, not forms

Survey Structure in AppGram

Understanding the hierarchy helps you communicate effectively with Claude:

Survey
├── Node (Question 1)
│   ├── answer_yes → Node 2
│   └── answer_no → Node 3
├── Node (Question 2)
│   └── options: [Option A, Option B, Option C]
└── Node (Question 3)
    └── result_message: "Thank you for your feedback!"

Question Types

Creating Your First Survey

Step 1: Create the Survey Container

Ask Claude:

"Create a customer satisfaction survey called 'Q1 2024 Feedback' for my project"

Claude creates the survey:

{
  "project_id": "your-project-id",
  "name": "Q1 2024 Feedback",
  "slug": "q1-2024-feedback",
  "description": "Quarterly customer satisfaction survey"
}

Step 2: Add Questions (Nodes)

Ask Claude to design the questions:

"Add these questions to the survey: 1. NPS rating (0-10): How likely are you to recommend us? 2. Multiple choice: What's your primary use case? (Options: Personal, Small Business, Enterprise, Agency) 3. Checkboxes: Which features do you use most? (Dashboard, Reports, API, Integrations, Mobile App) 4. Paragraph: What could we improve? 5. Yes/No: Would you be interested in a case study interview?"

Claude creates each node with the appropriate configuration:

// NPS Question
{
  "survey_id": "...",
  "question": "How likely are you to recommend us to a friend or colleague?",
  "question_type": "rating",
  "min_rating": 0,
  "max_rating": 10,
  "is_required": true
}

// Multiple Choice
{
  "survey_id": "...",
  "question": "What's your primary use case?",
  "question_type": "multiple_choice",
  "options": "[{\"value\":\"personal\",\"label\":\"Personal\"},{\"value\":\"small_business\",\"label\":\"Small Business\"},{\"value\":\"enterprise\",\"label\":\"Enterprise\"},{\"value\":\"agency\",\"label\":\"Agency\"}]",
  "is_required": true
}

// Checkboxes
{
  "survey_id": "...",
  "question": "Which features do you use most? (Select all that apply)",
  "question_type": "checkboxes",
  "options": "[{\"value\":\"dashboard\",\"label\":\"Dashboard\"},{\"value\":\"reports\",\"label\":\"Reports\"},{\"value\":\"api\",\"label\":\"API\"},{\"value\":\"integrations\",\"label\":\"Integrations\"},{\"value\":\"mobile\",\"label\":\"Mobile App\"}]"
}

Building Branching Surveys

The real power comes from conditional logic. Create surveys that adapt based on responses.

Example: NPS Follow-up Flow

Describe the flow to Claude:

"Create an NPS survey with branching: - Ask NPS score (0-10) - If score is 9-10 (promoter): Ask 'What do you love most about us?' then 'Would you write a review?' - If score is 7-8 (passive): Ask 'What would make you rate us higher?' - If score is 0-6 (detractor): Ask 'What disappointed you?' then 'Can we contact you to make this right?'"

Claude creates the branching structure:

// Root: NPS Question
{
  "question": "How likely are you to recommend us? (0-10)",
  "question_type": "rating",
  "min_rating": 0,
  "max_rating": 10
}

// Branch: Promoter (9-10)
{
  "question": "What do you love most about us?",
  "question_type": "paragraph",
  "parent_id": "nps-node-id"
}

// Branch: Detractor (0-6)  
{
  "question": "We're sorry to hear that. What disappointed you?",
  "question_type": "paragraph",
  "parent_id": "nps-node-id"
}

// Detractor follow-up
{
  "question": "Can we contact you to make this right?",
  "question_type": "yes_no",
  "parent_id": "detractor-feedback-node-id",
  "result_message": "Thank you. We'll reach out within 24 hours."
}

Common Survey Patterns

1. Product-Market Fit Survey

"Create a PMF survey with: - 'How would you feel if you could no longer use our product?' (Very disappointed / Somewhat disappointed / Not disappointed) - 'What type of person would benefit most from our product?' - 'What is the main benefit you receive from our product?' - 'How can we improve the product for you?'"

2. Feature Request Survey

"Create a feature prioritization survey: - Multiple choice: 'Which area needs the most improvement?' (Performance, UI/UX, Integrations, Reporting, Mobile) - Checkboxes: 'Which upcoming features interest you most?' (List your roadmap items) - Paragraph: 'Describe a feature that would save you time' - Rating: 'How important is mobile access to you?' (1-5)"

3. Churn Prevention Survey

"Create an exit survey for canceling users: - Multiple choice: 'Primary reason for leaving?' (Too expensive, Missing features, Switched to competitor, No longer needed, Hard to use, Other) - If 'Too expensive': 'Would a discount change your mind?' (Yes/No) - If 'Missing features': 'What feature would have kept you?' - If 'Switched to competitor': 'Which competitor and why?' - Final: 'Any other feedback for us?'"

4. Onboarding Feedback Survey

"Create a day-7 onboarding survey: - Rating: 'How easy was it to get started?' (1-5) - Yes/No: 'Did you complete your first project?' - If No: 'What blocked you?' (Multiple choice: Confusing UI, Missing feature, Technical issue, Ran out of time, Other) - Multiple choice: 'How did you learn to use the product?' (Documentation, Video tutorials, Support chat, Trial and error) - Paragraph: 'What would have made onboarding easier?'"

Managing Surveys

Activating and Deactivating

"Activate the Q1 feedback survey"
"Deactivate the beta testing survey"
"List all active surveys"

Modifying Questions

"Add an option 'Other' to the use case question"
"Make the email question required"
"Change the rating scale from 1-5 to 1-10"
"Update the thank you message to include a discount code"

Reordering Questions

"Move the NPS question to the end of the survey"
"Reorder questions: use case first, then features, then NPS, then open feedback"

Analyzing Responses

Once responses come in, use Claude to analyze them:

View Responses

"Show me the last 20 responses to the Q1 survey"
"Get response details for response ID xyz"
"List all survey responses from this week"

Example Analysis Request

"Show me all responses to the NPS survey. Summarize: - Average NPS score - Common themes in promoter feedback - Main complaints from detractors - Suggested improvements"

Claude retrieves responses and provides analysis:

Pro Tips

1. Let Claude Design Your Survey

Instead of specifying every question, describe your goal:

"I want to understand why users aren't completing onboarding. Design a survey that identifies the main blockers, takes under 2 minutes, and ends with an offer to help."

2. A/B Test Question Wording

Create survey variants:

"Create two versions of the NPS survey: - Version A: Standard NPS question - Version B: 'On a scale of 0-10, how likely are you to tell a friend about us?'"

3. Segment by Response

"Show me only responses where NPS was 6 or below" "Find all responses mentioning 'pricing' in open feedback"

4. Export for Deep Analysis

"Export all survey responses as JSON for the data team"

Quick Reference: Survey Commands

Survey Management

"Create a survey called [name]"
"List all surveys"
"Get survey [id] with all questions"
"Activate/deactivate survey [id]"
"Delete survey [id]"

Question Management

"Add a [type] question: [text]"
"Update question [id] to be required"
"Add options [A, B, C] to question [id]"
"Set branching: if yes go to [node], if no go to [node]"
"Delete question [id]"

Response Management

"List responses for survey [id]"
"Get response [id] with answers"
"Show responses from the last 7 days"

Conclusion

Building surveys with Claude and AppGram MCP transforms user research from a tedious setup process into an intelligent conversation. You can:

• Design surveys by describing your research goals
• Build complex branching logic through natural language
• Iterate on questions without navigating forms
• Analyze responses with AI assistance

Start with a simple survey, then experiment with branching flows as you get comfortable. The conversational approach makes it easy to refine your surveys based on what you learn.

Next step: Connect AppGram MCP to Claude and try: "Create a quick 3-question feedback survey for my app"