Design Scenarios

A strong simulation starts with strong scenario design. Your goal is to represent real user conversations and define clear expectations for each step.

What a Scenario Contains

A conversation simulation scenario includes:

• Scenario Name
• Category
• One or more Conversation Steps
• For each step:
• User Message
• Expected Response
• Evaluation Criteria

The more realistic and measurable your scenario, the more useful your results.

Step-by-Step: Build a Scenario

1) Open Conversation Simulation

1. Go to Simulation -> Conversation.
2. Click New Conversation.

2) Add conversation steps

1. In Conversation Steps, define the user message for Step 1.
2. Fill the expected AI response for Step 1.
3. Add more steps with Add Message to simulate a real multi-turn conversation.

3) Generate and refine criteria

1. For each step, click Suggest/Rewrite Criteria.
2. Review generated criteria carefully.
3. Add manual criteria if needed.
4. Rewrite unclear criteria until they are specific.

4) Complete conversation info

1. Click Save in create mode.
2. The system can suggest a scenario name automatically.
3. Confirm or edit the name.
4. Select the category.
5. Confirm to create the scenario.

Criteria Writing Rules

Good criteria are testable. Avoid vague wording.

Use this pattern:

• Include required facts
• Include required policy boundaries
• Include prohibited behavior

Scenario Design Best Practices

1. Start from real conversation logs and support FAQs.
2. Keep one primary objective per scenario.
3. Separate happy-path and edge-case scenarios.
4. Include at least one escalation-trigger scenario if your workflow needs handoff.
5. Add model-sensitive scenarios (long reasoning, policy nuance, multilingual requests) for model-switch testing.

Ready for Execution

Once scenarios are ready, run and monitor them:

Run and Track