Skip to content

Prepare your data once.

Features · Datasets

Your filters, partitions, targets, and features carry across every model you build on it.

A dataset workspace showing the driver_age column's Claim Frequency distribution, a histogram by age band with a Claim Frequency line over it, beside a sidebar listing the dataset's properties, filters, partitions, targets, associated geography, and models. Toggling a filter, partition, or target in the sidebar updates the chart.
1 filter
1 target

Claim Frequency distribution

Upload

Import a CSV or Parquet file.

Imports run in the background so you can keep working.

A new dataset form with a name field, a source type picker showing Parquet, CSV, and a disabled Google BigQuery option, and a drag-and-drop area for the data file.
New dataset

Upload a data file

Correlation analysis

Spot the relationships in your data.

Brush a range in one column to see how it distributes across another.

A correlation analysis view with two count histograms, vehicle_power and driver_age, sharing a Count axis. Drag to brush a range in either histogram and the other updates to show how those rows distribute, with the selection highlighted in blue.

Correlation analysis

Count

Count

Filters

Filter with expressions.

Write a filter with functions and operators, then reuse it across models.

An editable expression builder for the filter Young Drivers, pre-filled with driver_age < 25. Type a column, operator, or number to edit it via autocomplete.
New filter
driver_age<25

Partitions

Partition by time, region, or random fold.

Hold out by time or region, carve a test set, build cross-validation folds, and more.

An editable partition builder named Policy Years with two parts, each an expression: 2024 where policy_year = 2024, and 2025 where policy_year = 2025. Parts can be added, removed, and edited.
New partition
policy_year=2024
policy_year=2025

Branches

Nothing changes underneath your models.

Create branches to experiment without affecting your previous work, then merge your changes through a review when you're ready.

A datasets table listing Personal Auto, Homeowners, and Commercial Auto Policies, all Ready, each with a branch count that expands to show its branches. A search field and a New dataset button sit above the table.

Datasets

Dataset name
Branches
Created
Status

Personal Auto

Jun 4, 2026

Ready
DefaultDefault
q1-refresh

Homeowners

May 28, 2026

Ready

Commercial Auto Policies

Jun 5, 2026

Ready

More features

Explore the rest.

Try Prediction Lab.

Install the desktop app, or book a 15-minute demo.

Download