Overview
A run plan is a reusable template that defines a comprehensive simulation configuration. Instead of running individual simulations one by one, run plans automatically generate test matrices by combining scenarios, personas, agent endpoints, and evaluators - making it easy to execute systematic testing manually or via schedule.Navigate to Simulations → Run Plans in your Roark dashboard to create and manage run plans.

What is a Run Plan?
Think of a run plan as a test suite generator. It creates all possible combinations of your selected components:The Matrix Formula
- 2 scenarios (billing inquiry, technical support)
- 3 personas (happy customer, frustrated customer, confused elder)
- 2 agent endpoints (production agent, staging agent)
- 1 iteration each
Creating a Run Plan
Run plans are created through a step-by-step wizard that guides you through selecting each component:Configuration Steps
1
Select Scenarios
Choose one or more conversation scenarios that define what will be tested
2
Select Personas
Pick the customer personas to use - each brings different voice characteristics and behaviors
3
Choose Agent Endpoints
Select which agents to test against (see inbound vs outbound section below)
4
Add Evaluators
Choose evaluation criteria to assess each simulation
5
Review & Configure
Set iterations, concurrency limits, and preview the test matrix before creation
Configuration Options
| Setting | Description | Default |
|---|---|---|
| Name | Descriptive name for your run plan | Required |
| Description | Optional details about the test purpose | None |
| Direction | Inbound or Outbound testing mode | Inbound |
| Iterations | How many times to repeat each combination | 1 |
| Max Concurrent Jobs | Parallel execution limit | 10 |
Testing Direction
Run plans can be configured for either inbound or outbound testing, which determines who initiates the calls:- Inbound: Roark calls your agent (immediate execution, multiple iterations)
- Outbound: Your agent calls Roark (async execution, single iteration)
Learn more about choosing the right direction in our Inbound vs Outbound guide.
Understanding the Test Matrix
The power of run plans lies in their ability to systematically test all combinations:Example Matrix
Let’s say you’re testing a customer service agent with: Components Selected:- Scenarios: Password Reset, Billing Inquiry
- Personas: Calm Professional, Frustrated Customer
- Agent Endpoints: Production Agent
- Evaluators: Task Completion, Customer Satisfaction
| # | Scenario | Persona | Agent | Tests |
|---|---|---|---|---|
| 1 | Password Reset | Calm Professional | Production | Task Completion, Satisfaction |
| 2 | Password Reset | Frustrated Customer | Production | Task Completion, Satisfaction |
| 3 | Billing Inquiry | Calm Professional | Production | Task Completion, Satisfaction |
| 4 | Billing Inquiry | Frustrated Customer | Production | Task Completion, Satisfaction |
Previewing Your Run Plan
Before executing, you can preview the complete test matrix:
The preview shows:
- Total number of test cases
- All combinations that will be generated
- Estimated execution time
- Resource requirements
Executing Run Plans
Run plans can be executed in two ways:Manual Execution
Click the “Run” button on any run plan to execute immediately. Process:- System generates all test case combinations
- Simulations execute according to concurrency limits
- Each simulation is evaluated
- Results aggregate in real-time
- Complete report available when finished
Scheduled Execution
Attach run plans to schedules for automated testing:Best Practices
Start Small
Start Small
Begin with a few combinations to validate your setup before scaling up
Use Meaningful Names
Use Meaningful Names
Name run plans descriptively (e.g., “Daily Production Regression Suite”)
Balance Coverage
Balance Coverage
Include both happy path and edge case scenarios in your matrix
Set Appropriate Concurrency
Set Appropriate Concurrency
Match concurrency limits to your agent’s capacity to avoid overload
Review Failed Cases
Review Failed Cases
Analyze failures to identify patterns and improve agent responses
Common Use Cases
Regression Testing
Create a comprehensive run plan with all critical scenarios and run it before each deployment.A/B Testing
Set up identical run plans against different agent versions to compare performance.Load Testing
Use high concurrency and multiple iterations to stress-test your infrastructure.Compliance Validation
Include compliance-focused scenarios and evaluators to ensure regulatory adherence.Integration with Other Features
Run plans work seamlessly with other Roark features:- Scenarios - Define the conversations
- Personas - Provide voice variety
- Evaluators - Assess quality
- Schedules - Automate execution
- Reports - Analyze results