If you have results from previous experiments — whether from manual testing, classical DoE, or a prior optimization run — you can feed them into an ongoing experiment. The optimizer uses this data to skip or shorten the initial exploration phase and start making informed recommendations sooner.
Where to Find It
Open the experiment you want to enrich, then go to the Historical Data section in the experiment configuration. This section lets you attach one or more datasets to the experiment.
Adding a Dataset
You have two ways to add a dataset:
Upload a CSV file — drag and drop a file onto the drop zone, or click it to open a file picker. CSV and TXT files up to 5 MB are accepted. The file is parsed immediately.
Browse existing datasets — click Browse datasets to pick a dataset already in the platform, including datasets generated by other experiments.
You can attach multiple datasets at once. Each one appears as its own row in the list with a status badge — Incomplete (orange) while its mapping still needs work, or Ready (teal) once it is fully configured. All datasets must reach Ready before the section can be saved.
Configuring the Column Mapping
After adding datasets, click Configure datasets to open the mapping editor. The editor lays out your attached datasets as side-by-side columns and your experiment's variables and results as rows, so you can map every dataset in one place and compare them at a glance.
For each variable or result in the experiment, you choose how each attached dataset fills it in:
Map a column — select an existing column from the dataset. The optimizer reads values from that column for each data point.
Set a constant — enter a fixed value that applies to every row in the dataset. Use this when a parameter was held fixed during data collection and is not present as a column.
Leave unmapped — skip the row. The dataset will not provide values for that parameter or measurement.
Each attached dataset must be fully mapped before the configuration can be saved. A Required label sits next to the Configure datasets button until every dataset is ready, mirroring the per-dataset Ready / Incomplete badges in the list.
What Happens After You Save
Once saved, the attached datasets are included in the optimizer's training data from the next iteration. The more relevant and well-spread the historical data is, the faster the optimizer can identify promising regions of the parameter space.
Limitations
If a dataset was generated by an SDLabs experiment (an output dataset), constants cannot be used when mapping it. Only column mapping is supported. See Attaching an Experiment Output Dataset: Live Link and Mapping Limitations for details.
Tips
Match the conditions — historical data should come from the same or very similar experimental setup. Data from a different scale, protocol, or instrument may not transfer well.
Cover a wide range — broad coverage across the parameter space gives the model a better starting map than data clustered in one region.
More is generally better — but 10 high-quality, well-spread points are worth more than 100 noisy points in a narrow region.