Analyze
Create Analysis
A three-step wizard walks you from picking a dataset to running a statistical analysis and reading the results — no separate query or script required.
Quick Start
- 1
Choose a dataset
Pick a table your organization has in the lakehouse from the searchable picker, or enter a table name manually if it isn't listed yet.
- 2
Select columns
Check the columns to analyze from the table's schema. Analyses run on numeric columns — non-numeric ones are either ignored or reported as non-numeric depending on the method.
- 3
Pick an analysis type
Choose from descriptive statistics, correlation, a t-test, regression, chi-square, ANOVA, or distribution fitting. The fields shown adjust to match what that method needs.
- 4
Run it
The analysis submits as a job and the page polls until it completes — results render inline once it's done.
- 5
Iterate
Re-run the same analysis, or start a new one from scratch without losing your place.
Choosing a dataset
Table pickerSearchable list of your tenant's tables across ingestion, transforms, and dbt schemas — shown as a badge next to each table name.Manual entryFalls back automatically if your dataset list fails to load, or if you have no tables yet. You can also switch to it manually if a table isn't synced into the picker.Schema suffixingestion, transforms, or dbt — tells the engine which schema to read the table from.Analysis types
The fields shown in step 2 change based on what you pick. See Statistical Methods for details on each one.
Descriptive statisticsRuns over the selected columns as-is — no extra fields needed.Pearson correlationAlso runs over the selected columns as-is.Two-sample t-testRequires Column A and Column B (must be different). Optionally assume equal variance — off by default, which runs Welch's t-test instead of Student's.Linear regression (OLS)Requires a dependent variable (Y) and one or more independent variables (X). Predictors must be numeric.Chi-square test of independenceRequires two categorical columns — works on text or coded categories.One-way ANOVARequires a numeric value column and a categorical group column — compares the value column's mean across the group's levels.Distribution fittingRequires a numeric column and a distribution family (Normal, Log-normal, or Exponential). Reports a Kolmogorov–Smirnov goodness-of-fit plus a normality test.- The engine only reads your own tenant's data — you can't accidentally point an analysis at another tenant's table.
- You can't advance past a step until it's valid — e.g. picking two different columns for a t-test, or at least one predictor for regression.
- A failed job shows the specific error message instead of a generic failure — check it before re-running.
- Re-run keeps your exact configuration; New analysis clears everything and starts over.