Automatic optimization

When your aim is simply to find the best configuration, choose the Optimizer design. Instead of running a fixed set of points, it searches: it learns from every run and decides where to look next, so it can find a strong configuration in far fewer runs than trying everything.

How it searches

The optimizer picks its method to fit your experiment, and you never have to choose one. When the factors are mostly numeric, it builds a model of how the result responds and heads toward the most promising area, a run at a time. When the space is large and full of options rather than numbers, it switches to an evolutionary search that breeds better configurations across generations. The panel that tracks the search explains, in plain language, what it is doing and why.

Budget and stopping

The optimizer always runs against a budget you set, in number of runs. It stops when the budget is spent, or earlier when results stop improving, so you are never left watching a search that has nothing more to find. As with any experiment, you can stop and keep results at any point, because every finished configuration is already recorded.

The right number of replications

Because a stochastic model gives a slightly different answer each run, the optimizer runs each configuration enough times to tell a real difference from noise, rather than making you guess a replication count up front. That puts tighter confidence where it counts, without wasting runs on configurations that are clearly out of contention.

More than one goal

When you care about more than one result, some configurations trade one goal against another: more throughput, but a longer wait. The trade-off view plots them and highlights the frontier, the configurations that nothing else beats on every goal at once. Points that are beaten on all goals are greyed out. Pick any frontier configuration and Save as Scenario to keep it. The frontier is judged over all your goals, not only the two drawn on the chart.

Warm start

You do not have to begin cold. A new experiment can warm start from the best configurations of a previous one, so a follow-up study picks up where the last left off instead of rediscovering what you already know.