Assignments
Just as genes are assigned during reproduction, our framework uses Assignments - the critical linkage mechanism that connects Units to specific Variants for controlled testing and measurement.
What
Just as genes are assigned during reproduction, our framework uses `Assignments` - the critical linkage mechanism that connects Units to specific Variants for controlled testing and measurement.
## What is an Assignment?
An Assignment is the binding relationship between a Unit and an Experiment Variant:
- **Deterministic**: Creates a fixed Unit-Variant relationship
- **Persistent**: Maintains consistency throughout the Unit lifecycle
- **Traceable**: Provides complete context for measurement analysis
- **Randomized**: Distributed according to specified weights for statistical validity
## Key Characteristics
1. **Assignment Mechanics**
- Created at Unit birth/initialization
- One Assignment per active Experiment per Unit
- Random distribution based on Variant weights
- Sticky persistence throughout Unit lifecycle
- Cannot be changed once established (ensures data integrity)
2. **Technical Implementation**
- Stored in cookies for web applications
- Encrypted in local storage for mobile apps
- Persisted in database for server-side reference
- Synced across devices for authenticated users
- Gracefully handles edge cases (cookie clearing, multiple devices)
3. **Assignment Context**
- Records timestamp of assignment creation
- Maintains reference to originating Experiment
- Links to specific Variant configuration
- Associates with Unit characteristics for segment analysis
- Preserves version information for reproducibility
4. **Analysis Framework**
- Enables cohort segmentation by Assignment groups
- Provides filtering mechanism for metric comparison
- Facilitates statistical significance testing
- Supports multivariate analysis across multiple Assignments
- Enables interaction effect analysis between Experiments
## Evolutionary Intelligence
Assignments create the inheritance mechanism for product evolution:
1. Consistent Assignments ensure valid experimental data
2. Assignment distributions control statistical power
3. Assignment permanence prevents contamination effects
4. Multiple concurrent Assignments enable complex testing scenarios
5. Assignment metadata supports rich analytical contexts
The ultimate goal is reliable, statistically valid connections between Units and test Variants, creating a robust foundation for experimental validation and continuous product improvement.By Eduarda Ferreira