Skip to main content

​
Overview

Calibration is the process of adjusting your model’s parameters so that its outputs match observed real-world data or expected behavior. In Veydra, calibration happens in Calibrate mode — a dedicated workspace with three approaches: manual tuning, automated optimization, and data-driven fitting.
Parameter changes in Calibrate mode apply to both the baseline and current scenario. This is different from Experiment mode, where changes only affect the current scenario.

​
Calibration Panel

When you switch to Calibrate mode, a dedicated Calibration Panel opens on the left side of the playground. It offers three tabs:

Manual

Adjust parameters by hand and observe effects in real time

Auto

Automated optimization using algorithms (coming soon)

Data

Upload observed data and calibrate against it (coming soon)

​
Manual Calibration

Manual calibration gives you direct control over every parameter in your model. This is the recommended starting point — it builds intuition about which parameters matter most and how they influence model behavior.

​
How It Works

  1. Open Calibrate mode — Click the Calibrate button in the top navigation bar
  2. Expand a submodel section — Parameters are grouped by submodel; click any section to expand it
  3. Adjust a parameter — Use the slider or type a value directly
  4. Run the model — See how the output changes in the chart area
  5. Compare against baseline — Use Compare Mode to see before/after

​
What You See

  • Parameter sliders organized by submodel (all sections start collapsed so you can navigate to the group you need)
  • Charts showing real-time simulation output as you adjust values
  • Stock-Flow Diagram available for reference (minimized by default)

​
Tips for Effective Manual Calibration

Focus on parameters that have the largest impact on model behavior. Stocks’ initial values and key flow rates typically matter most.
Keep Compare Mode enabled to see how your current parameter values differ from the baseline. This makes it easy to spot improvements and regressions.
Expand one parameter group, make adjustments, observe the effect, then move on. This avoids confusion about which change caused which effect.
Use the reset button to return parameters to their defaults. This gives you a clean starting point without losing your scenario history.

​
Automated Calibration (Coming Soon)

Automated calibration is not yet enabled. Contact sales to learn about activating this capability for your organization.
Automated calibration uses optimization algorithms to find the best-fit parameter values. Instead of adjusting sliders by hand, you define which stocks to optimize and let the system search for optimal values.

​
Planned Features

  • Optimization methods — Genetic Algorithm, Gradient Descent, and Bayesian Optimization
  • Stock selection — Choose which model stocks to include in the optimization objective
  • Iteration control — Set maximum iterations and convergence criteria
  • Progress tracking — Monitor optimization progress in real time

​
Typical Workflow

1

Select method

Choose an optimization algorithm based on your model’s characteristics
2

Configure stocks

Select which stocks the optimizer should focus on
3

Run optimization

The system iterates through parameter combinations to minimize error
4

Review results

Compare the optimized output against your target data
Start with manual calibration to understand your model’s sensitivity, then use automated tools for fine-tuning. Manual exploration will help you set reasonable bounds for the optimizer.

​
Data-Driven Calibration (Coming Soon)

Data-driven calibration is not yet enabled. Contact sales to learn about activating this capability for your organization.
Data-driven calibration lets you upload historical or observed data and automatically fit your model’s parameters to match it. This is the most rigorous approach — especially useful when you have time-series data from the real system you’re modeling.

​
Planned Features

  • CSV / Excel / JSON upload — Bring your own time-series data
  • Automatic column mapping — Match data columns to model stocks
  • Fit quality metrics — RMSE, R², and visual overlay of model vs. data
  • Iterative refinement — Adjust the fit and re-run as needed

​
Calibration vs. Experimentation

Understanding the difference between Calibrate and Experiment modes is key to using Veydra effectively:
Calibrate ModeExperiment Mode
PurposeMatch the model to realityExplore what-if scenarios
Parameter changes affectBoth baseline and current scenarioCurrent scenario only
BaselineBeing definedFixed (from calibration)
Typical question”What parameter values make the model match historical data?""What happens if we double the growth rate?”
The recommended workflow is: Design → Calibrate → Experiment → Decide.

​
Next Steps

Playground Overview

Return to the full playground documentation

Model Controls

Learn about parameter sliders, activation, and scenario management

Behavior Analysis

Analyze time-series patterns to validate your calibration

Experiment Mode

After calibrating, explore scenarios in Experiment mode