When you have a mathematical model of your process and need to find the exact factor settings that deliver the best possible outcome, use this method. Whether you want to maximize a yield, minimize a defect rate, or hit a specific target value, it identifies the precise conditions you need.
Finding the Ideal Process Conditions
Desirability function optimization is a practical way to translate complex process requirements into clear, actionable data. By defining whether you want a value to be as high as possible, as low as possible, or centered on a specific target, you provide the framework needed to navigate your factor space.
The method systematically evaluates different combinations of your inputs to find the point where the desired outcome is most likely to occur. It moves beyond simple observation, using your data to mathematically calculate the most efficient path forward.
Handling Target Goals
You can define your objective in three distinct ways. 'Max target' focuses on increasing output, 'min target' focuses on reduction, and 'target best' allows you to aim for a precise value within a specified range. This flexibility ensures that the results align with your actual operational priorities.
The tool checks the model across a wide range of your factors, ensuring the recommended settings remain within the boundaries of your established test space. This prevents the calculation of 'optimal' points that fall outside the realm of what you have physically tested.
Interpreting the Results
Once the calculation is complete, Lattice provides the optimal settings in both coded and natural units. You will also see a predicted response value and a desirability score, which acts as a confidence rating for how well the optimal point meets your criteria.
If the desirability score is low, it often suggests that your current process limits may be too restrictive. In these cases, you might consider adjusting your input ranges to allow the tool more flexibility to find a higher-scoring configuration.
1 · Intent → method
An LLM picks optimize_desirability from a fixed catalog.
2 · Method → numbers
Deterministic Python engine runs the math. Same input → same output.
3 · Numbers → plain language
A second LLM translates the result into your domain’s vocabulary.
How does this method decide what counts as 'optimal'?
It converts your process results into a desirability score between 0 and 1. The optimization tool then adjusts your input settings to find the combination that results in the highest score, balancing your specific goals for the outcome.
What happens if the model is not very accurate?
If the model's reliability—measured by its R-squared value—is low, the tool provides a warning. This helps you understand that the resulting optimal settings should be treated with caution, as the underlying predictions may not be precise enough to rely on.
Tool input schema
Schema for optimize_desirability not exported yet (run pnpm export:registry).