Causal inference: Data tab

Upload a CSV or JSON file, check it for ID-like columns, create a synthetic sample, and see the datasets and storage your plan allows.

What the tab is for

The Data tab holds the tables causal discovery runs on: the ones you upload, synthetic samples you generate, and five built-in example datasets. From top to bottom it shows your plan's storage, the upload area, the sample generator and the list of datasets.

The plan card

The card at the top shows your plan's name, how many datasets you have used out of how many, the storage used out of the total, Datasets Left and Max File Size. The bar turns amber above 70 % and red above 90 %. Delete datasets you no longer need to free space (each dataset in the list has a delete button). The limits of each plan are in Billing and plan limits (signed in).

Upload your own dataset

Select Click to upload and choose a .csv or .json file.
The page reads the file in your browser and shows a Data Preview of the first 10 rows, the number of rows and columns and the file size.
Check Dataset Name (it starts as the file name) and choose a Domain. General Purpose applies no domain knowledge; a named domain shows its typical variables and its root causes, leaf effects, required and forbidden edges. The domain starts on General Purpose; pick a named domain only when your data is about it.
Select Upload Dataset. A bar shows the progress; the dataset then appears under Available Datasets.

A CSV file selected: the file name, size and row count on the left, the dataset name and domain fields below, and the data preview on the right — The Sachs protein data (853 rows) ready to upload

What the file must look like.

A .csv file: a header row with one name per column, then one row per observation, separated by commas. Every column is used as a variable.
Numbers only, in every cell. When a column has empty cells or text (names, categories, dates), the page lists it under Columns that are not numbers with how many cells are affected, and Upload Dataset stays disabled until you select Leave these columns out or fix the file and choose it again. No cell is ever uploaded as 0 in place of a missing value.
A .json file works too if it is an array of objects with the same keys (one object per row).
Excel workbooks are not read: save the sheet as CSV first. The file picker offers CSV and JSON only.
The size limit per file is your plan's Max File Size, also shown under the upload area.

ID-like columns

When a file is read, the page looks for columns that identify rows rather than measure something (row numbers, codes, unique keys, timestamps, constant columns). If it finds any, the ID Column Detection dialog opens.

The ID Column Detection dialog listing the column sample_id with a HIGH risk badge, its unique ratio and the reasons it was flagged — The check flags a row-number column added to the Sachs data

Each flagged column shows a risk level (HIGH, MEDIUM, LOW), the share of unique values, its strongest correlation with another column and the reasons.
Tick the columns to drop and select Remove … Column(s) & Continue, or Keep All Columns. Cancel closes the dialog and keeps the choice open: the … ID Column(s) Detected button above the preview reopens it.
With nothing flagged, the preview says Dataset is Causal-Ready - No ID columns detected.

Create a sample dataset

Create Sample Dataset asks the server to generate synthetic data for a domain: give it a name, pick a domain and a Sample Count (10 to 10,000 rows), then Create Dataset. Use it to try the workflow; its values are generated, not measured.

The Create Sample Dataset card: dataset name, domain with its variables and constraints, and sample count — Create Sample Dataset

Available datasets

The list shows the five built-in example datasets (Cardiovascular Health, Stock Market Dynamics, Climate Change Factors, E-Commerce Conversion, Manufacturing Quality) and your organization's own, each with its domain, row count and variables.

The eye button opens Dataset Preview: the variables and up to the first 100 rows.
The bin button asks first; Delete dataset deletes the dataset and its stored file, and it cannot be restored. Graphs already discovered from it stay. Deleting a built-in example is refused.
The refresh button reloads the list.

Your uploads are stored for your organization only, encrypted at rest. The download button next to the eye saves a dataset again as CSV: the rows as stored (after any columns you left out), under its variable names.

Common messages

Message	What it means
Please upload a CSV or JSON file. Excel workbooks are not read …	The file is of another type. Save an Excel sheet as CSV first.
Every column must hold numbers. Not numeric or empty: …	A kept column has empty or text cells; the message names each column and how many cells. Leave those columns out or fix the file.
Invalid JSON format	The `.json` file could not be read as JSON.
Please select a file and enter a dataset name	The name field is empty.
You have reached your dataset limit (…)	Your plan's number of datasets is used up. Delete one or change plan.
File size (…) exceeds maximum allowed (…)	The file is larger than your plan allows per dataset.
Insufficient storage. You have … remaining.	The file does not fit in your plan's remaining storage.
Upload failed: …	The server refused or could not store the dataset; the rest of the message says why.