Datasets overview

The Plans page is a model’s workspace overview: a table of every dataset submitted to that model, with the actions to submit new ones, and the tools to search, filter, and categorize them for auditability.

Find it by opening a model from the top-navigation Models catalog, then selecting Plans.

What a plan represents

Each row in the table is a dataset: one solve operation submitted to the model, together with its resulting scores and metrics. A plan can be a first solve of a new planning problem, a re-solve of an existing dataset, or a revision created from a patch. See Dataset lifecycle for the statuses a dataset moves through, and Dataset details for how to interpret its scores and metrics in detail.

Table columns

The table lists the following columns for each dataset:

Name: The descriptive name given to the dataset.
Configuration profile: The configuration profile used to solve the dataset.
Tags: Any labels attached to the dataset. See Dataset names and tags below for how to use tags to organize and filter datasets.
Hard, Medium, Soft: The dataset’s constraint scores.
Output metrics defined by the model.

The exact set of output metric columns, depends on the model. Each model defines its own metrics, as explained in Dataset details.

You can customize which columns are displayed, including additional metrics defined by the model.

Plans versus Revisions

The page has two tabs:

Plans: Shows one row per planning problem you’ve submitted, that is, one row per chain of related solves. If a plan has been revised, for example through a re-solve or a patch, its row reflects the latest revision’s scores and metrics, not the original submission’s.
Revisions: Shows every dataset, including revisions created from a re-solve or from a patch.

Use the Plans tab to see one entry per planning problem you’ve submitted, always reflecting its most recent revision. Use the Revisions tab to see the full history of changes applied to a plan over time, for example through the patch endpoint.

Creating a new plan

Click New plan to submit a new dataset to the model. You can specify the input data, an optional name and tags, the configuration profile to use, and any configuration overrides for this submission.

Dataset names and tags

Our platform provides two ways to add metadata to datasets:

Name: Each dataset can be assigned a descriptive name.
Tags: Datasets can be labeled with one or more tags to facilitate filtering and organization.

Both the name and tags can be provided when submitting a dataset (via the Platform UI or as part of the json file) or edited later on the Plans Overview page in the Platform UI.

Searching and filtering datasets

You can search for datasets by name using the search bar in Plans Overview.

Click Add filter to drill down further:

ID filtering: Find a specific dataset by filtering on its unique dataset ID.
Tag filtering: Only display datasets that have a certain tag, or exclude a certain tag from the results.
Status filtering: Only display datasets in a specific status. See the Dataset lifecycle for a complete overview of all possible statuses.
Configuration profile filtering: Only display datasets that ran with a certain configuration profile.
Date range filtering: To only show datasets from a certain date range, filter on the field "Started at".
Deleted datasets: Any dataset that is deleted is kept in Trash for a certain amount of time. Filter on "Deleted: Yes"
Created from filtering: Filter datasets based on how they were created:
- Request: Created from a new API call.
- Input: Created from an existing dataset input (for example, a re-solve).
- Patch: Created from a patch update (see Dataset revisions with /from-patch).
Model version filtering: Filter datasets based on the model version used to generate them, including:
- Model SDK version
- Model version
- Model build time
- Model build branch
- Model branch

It’s possible to combine multiple filters.

On the Plans Overview page, you can choose which columns to show for each dataset, including any of the metrics defined by the model. This allows you to analyze the evolution of these metrics over time or quickly spot anomalies within the filtered datasets.

Exporting search results

Click Export CSV on Plans Overview to download the datasets that match your current search and filters.

The export includes the latest 100 matching datasets. If more datasets match your search and filters, this cap is shown in a tooltip on the button and in a toast after the export completes.

Best practices for using tags

We recommend using tags to:

Segment your data: Assign different tags to represent distinct segments in your data, like regions or departments. For example, give each region you plan in a separate tag identifying that region. This allows you to later search for and compare all plans in a specific region efficiently.
Distinguish planning types: Use tags to differentiate between nightly planning, real-time planning and reference plans. This helps track how often real-time plan adjustments are necessary or to compare nightly planning with actual executions.
Separate simulations from operational plans: Clearly mark simulation datasets (e.g. goal alignment experiments or test scenarios) separately from datasets to be used in actual operations. This ensures that test results don’t interfere with live planning data.
Enable meaningful trend analysis: A consistent tagging strategy lets you filter by tag on the Insights page, so trends reflect one business unit, region, or planning type instead of mixing them together.

To keep datasets of your production systems distinct from development or staging environments, we recommend using different tenants and discourage using tags. This ensures a clear separation of environments, preventing test or experimental data from affecting production operations.

By consistently categorizing your datasets using meaningful names and well-structured tags, you can streamline your workflow and make data-driven decisions more effectively.