After each recommendation cycle, the optimizer produces a batch of recommendations — specific parameter combinations it wants you to test. This article explains what recommendations contain, how to submit your results, and what options you have along the way.
What Are Recommendations
Each recommendation is a row of parameter values that the optimizer has selected. If your experiment has three parameters (Temperature, Solvent, Concentration), each recommendation specifies a value for all three.
The number of recommendations per iteration equals your parallelization setting — the number of simultaneous measurements you can perform, which you set during experiment setup.
How Predictions Work
Alongside the recommended parameter values, the optimizer also shows predicted measurement values — what it expects you will measure for each recommendation. Predictions include an uncertainty range (68.3% confidence interval), shown when you hover over a predicted value.
During initial exploration, predictions may not yet be available because the model is not yet available. Once guided optimization begins, predictions become increasingly accurate as more data is collected.
If your experiment has multiple objectives with a scalar strategy (Weighted Sum or Hierarchy), a Merit column also appears. Merit is a normalized score between 0 and 1, where 1 represents the best result observed so far.
Submitting Measurements
After running the recommended experiments in your lab, enter the measured values into the corresponding columns in the recommendations table. Then click Submit results.
Submitting results triggers a new recommendation cycle — the algorithm analyzes all available data and produces the next batch of recommendations. You will see the "Recommending" state while this is in progress.
Adding Extra Datapoints
You are not limited to the optimizer's suggestions. Click Add datapoint to add a row with your own parameter values. This is useful when:
You already know a promising region and want to include it in the optimization data.
You ran additional experiments outside the optimizer's recommendations.
You want to test a specific hypothesis.
Extra datapoints are treated the same as optimizer-suggested ones during optimization.
Uploading Results via CSV
For large batches, you can upload measurement results from a CSV file instead of entering them manually. Click the upload icon in the recommendations toolbar and select your file. The CSV columns must match the measurement names defined in your experiment.
Downloading Recommendations
Click the download icon in the recommendations toolbar to export the current recommendations as a CSV file. This is helpful when you need to share the suggested experiments with a lab technician or feed them into an automated system.
Good to Know
You can only submit results when the status shows New recommendations. You cannot submit while the algorithm is generating recommendations.
Values outside the defined parameter range are flagged with a warning but can still be submitted.
Modifying experiment configuration (adding parameters, changing bounds) while waiting for measurements is possible but may affect the optimizer's model.