Skip to main content

​
When to Use

Use Follow-up Follower Simulator to test suggested pathways of follow-up questions. Essential after changes that affect follow-up generation. Ensures follow-up questions are appropriate and answerable.

​
Parameter Configuration

​
Core Parameters

You can seed the simulator with either specific questions or question topics: Initial Questions: Specific starting questions to begin conversations (can add multiple) 
Initial Questions: ["What's the best paint for my kitchen cabinets?", "How do I choose the right drill bit?"]
Initial Question Topics: Broader categories for starting conversations (can add multiple) 
Initial Question Topics: ["Cabinet refinishing guidance", "Power tool selection help"]
Selection Strategy: How follow-up questions are chosen from chatbot suggestions 
Selection Strategy: Random or Auto
Max Turns is a common parameter that can be configured for any simulator type.

​
Selection Strategy Options

Random Selection: Random selection chooses follow-up questions randomly from chatbot suggestions. This provides broad coverage of different conversation pathways and tests general follow-up quality across various directions. Auto Selection: Auto selection uses Blast’s model to intelligently select follow-up questions. It focuses on areas where the chatbot shows weakness or potential issues, targeting pathways most likely to reveal problems. This optimizes testing efficiency by concentrating on vulnerable areas.

​
Initial Question vs Initial Question Topic

Initial questions are best when you need to test specific scenarios, like validating a particular user journey or debugging a known issue. They allow you to measure follow-up quality from precise starting points. Initial question topics work better for broader testing. They let you explore general conversation patterns, test follow-ups across different question variations within a topic, and evaluate how well the chatbot guides users through topic-level discussions.

​
Configuration Examples

Specific Follow-up Testing: 
Simulator: Follow-up Follower
Initial Questions: ["What type of flooring is best for my basement?"]
Selection Strategy: Auto
Max Turns: 6
Number of Runs: 2
Total Conversations: 2 (1 initial question × 1 selection strategy × 2 runs) Topic-based Follow-up Exploration: 
Initial Question Topics: ["Deck building for beginners", "Kitchen renovation planning"]
Selection Strategy: Random
Max Turns: 8
Number of Runs: 1
Total Conversations: 2 (2 initial question topics × 1 selection strategy × 1 run)