> ## Documentation Index
> Fetch the complete documentation index at: https://gtm-resouces.getmaxim.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# How can I Evaluate Prompts Across Various Scenarios?

> Prompts that work perfectly for one use case often fail spectacularly in others. A prompt optimized for technical documentation might produce inappropriate responses for customer support. Scenario-based evaluation ensures your prompts perform well across the full range of situations your users will encounter.
Comprehensive scenario testing reveals edge cases, identifies failure modes, and validates that your prompt handles the diversity of real-world usage.

### Defining Evaluation Scenarios

Scenarios are specific situations or contexts where your prompt will be used. Well-designed scenarios capture:

* **User Intent Variations**: Different goals users might have (seeking information, requesting help, making purchases, reporting problems).

* **Input Diversity**: Various ways users might phrase similar requests, from concise to verbose, technical to casual.

* **Context Differences**: Different background information, conversation states, or environmental factors affecting the interaction.

* **Edge Cases**: Unusual, ambiguous, or challenging situations that might break typical prompt behavior.

* **User Personas**: Different user types (experts vs. novices, friendly vs. frustrated, aligned vs. adversarial).

### Maxim AI’s Scenario Evaluation Features

Maxim AI enables comprehensive scenario-based prompt evaluation:

* **Scenario Management**: Organize and version test scenarios with rich metadata and categorization.

* **Batch Evaluation**: Run prompts against entire scenario suites automatically, executing hundreds of tests in parallel.

* **Scenario-Specific Metrics**: Define and track different success criteria for different scenario categories.

* **Comparative Views**: Compare how different prompt versions perform across the same scenarios.

* **Failure Clustering**: Automatically group similar failures to identify common issues across scenarios.

* **Scenario Analytics**: Visualize performance breakdowns by scenario type, difficulty, or other attributes.

**Continuous Testing**: Integrate scenario evaluation into CI/CD to catch regressions before deployment.

### Best Practices for Scenario-Based Evaluation

* **Start Broad, Then Deep**: Begin with diverse scenarios covering all use cases, then add depth within important categories.

* **Update Scenarios Continuously**: Add new scenarios based on production failures and user feedback.

* **Balance Coverage and Efficiency**: Maintain comprehensive coverage while keeping test execution time reasonable.

* **Version Scenario Suites**: Track how your scenario collection evolves alongside your prompt development.

* **Share Scenarios Across Team**: Use scenarios as communication tools to align on expected behavior.

* **Monitor Scenario Drift**: Track whether real-world usage patterns match your scenario distribution.

By systematically evaluating prompts across diverse scenarios, you build robust AI applications that handle the full complexity of real-world usage, maintaining quality and reliability even in challenging edge cases.