How-to

Generate your first dataset from a template

The fastest path from nothing to a scored, downloadable dataset. Pick a prebuilt industry schema, review the plan, generate, and read the quality report. About five minutes, start to finish.

CrossRow team 5 min read

CrossRow ships with 125 industry templates: banking, healthcare, e-commerce, telecom, and more. Each one is a complete, prebuilt generation plan you can run as-is or adapt. This walkthrough uses the E-commerce Platform template, but every template follows the same four steps.

Step 1 · Browse templates

Pick a starting point

From the sidebar, open Templates. Each card shows the tables it includes and the capabilities it exercises (geo-consistency, formulas, event streams, and so on). Filter by industry across the top, or by capability just below, to find one close to what you need.

The CrossRow Templates gallery, showing industry template cards filterable by industry and capability
Templates gallery: filter by industry or by the capability you want to see

Click Use Template on the one you want. CrossRow copies it into your Saved Plans as an editable plan and opens it for review.

Step 2 · Review the plan

See exactly what will be generated

The plan is where CrossRow shows its work before generating a single row. At a glance you get the table count, total rows, column count, and foreign-key relationships. The Generation Sources strip breaks down how each column is produced: semantic values, correlations, sequential keys, and FK references.

The plan review screen showing 6 tables, 26,520 rows, generation sources, and a per-column breakdown where state and country are marked consistent with city
The plan review screen: generation sources up top, then every column with its type, how it is produced, and its null rate

Read down the column list and you can see the relationships the template already knows about. customer_id is a sequential key. gender and loyalty_tier are weighted categoricals. And the address columns are correlated: state and country are both marked consistent with city, so a customer in a given city gets the right state, postal code, and country rather than three independently random values. Add a hint or a rule to any column here, or just accept the template’s defaults. CrossRow also validates the plan at this stage and flags issues like type mismatches before anything runs, so you approve a plan you actually understand.

Step 3 · Generate

Run it, straight to S3

Open Generate Data, choose your row counts and output, and start the run. Generation streams with live progress, so large datasets report as they go. When it finishes, the run appears in your history with its row count, elapsed time, and the part that matters: a quality score.

The Data Generation page showing a completed E-commerce Platform run: 980 rows, completed, a 100/100 score button, advisories, and SQL, Explore, and Files actions
A completed run: 980 rows to S3 in 8.4 seconds, scored 100/100, with the generated files ready to download

From here you can open the data in SQL, Explore it in the browser, or download the Files. But first, open the score.

Step 4 · Read the quality report

Confirm it before you ship it

Click the score button to open the Data Quality Report. It is computed from the rows that were actually generated, not the plan that was supposed to produce them. The headline is the overall score; below it, the report attributes everything to specific tables and columns.

The Data Quality Report modal: an overall score of 100.0, tables and row counts, a list of learned fixes applied automatically, a uniqueness warning, and per-table statistics
The quality report: an overall score, the fixes CrossRow applied automatically, and honest warnings about anything it could not fully satisfy

Two things worth reading every time. First, the learned fixes applied automatically: here, row-by-row detection was switched on for several derived columns so their formulas compute correctly. Second, the warnings: this run flags that categories.category_name could not hit 95% uniqueness because the plan only defines 30 distinct values for 80 rows. That is exactly the kind of honest, specific signal you want. Nothing is hidden behind a green checkmark; the report tells you what held and what did not, so you can tighten the plan and regenerate if it matters.

That is the whole loop. Template → plan you can read → generation → a scored report attributed to specific columns. From here, download the files for a test fixture, wire the S3 output into a pipeline, or open Explore to inspect the data directly. Want to understand what the score actually measures? See how CrossRow scores its own output.

Run your first template now

125 industry schemas, ready to generate. No production data required.

Open CrossRow