When you need to reconcile competing goals—like maximizing yield while simultaneously minimizing defect rates—multi-response optimization finds the specific set of conditions that offers the best compromise. It mathematically balances your priorities, providing a single, clear recommendation that satisfies your requirements across all your different measured outcomes.
Balancing Complex Trade-offs
In many experimental environments, improving one result often comes at the cost of another. Multi-response optimization moves beyond simple single-goal analysis by mathematically integrating multiple outcomes into a unified objective.
By assigning specific desirability functions to each response, the method evaluates the 'goodness' of various combinations. This turns a complex search into a structured ranking of factor settings, highlighting the specific point where all your objectives perform at their collective best.
Prioritizing Your Objectives
Not all metrics are equal in every study. Multi-response optimization allows you to assign specific weights to each of your response variables. A higher weight signals that a particular outcome is a priority, pushing the solver to satisfy that requirement more aggressively within the final optimal configuration.
Whether you are targeting an exact value, maximizing output, or minimizing waste, the mathematical model accounts for these targets independently before aggregating them. This ensures the output reflects the strategic goals you have defined for your process.
Data-Driven Decision Making
When the optimization process concludes, you receive more than just a list of coordinates. The platform provides predicted values for each response at the recommended setting and the overall desirability score achieved.
This transparency lets you see exactly how well the final recommendation satisfies each of your goals. If the result shows a low desirability score, you can quickly revisit your target bounds or factor ranges to explore alternative process windows that might offer a more favorable trade-off.
1 · Intent → method
An LLM picks optimize_multi_response 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 multi-response optimization handle conflicting goals?
It uses a desirability function to convert each individual response into a score between 0 and 1. By calculating a weighted average of these scores, the method identifies the specific factor settings that provide the best overall performance compromise.
What happens if my constraints are impossible to meet?
If the optimization cannot find a solution that satisfies your input requirements, Lattice provides a warning rather than raising an error. This allows you to evaluate the current result, identify which goal is most difficult to reach, and adjust your target bounds accordingly.
Tool input schema
Schema for optimize_multi_response not exported yet (run pnpm export:registry).