Schema Drift Detector

What is Schema Drift Detector?

Schema Drift Detector compares two samples of the same dataset or API response and reports precisely what changed about its shape: fields added or removed, fields renamed, types changed, columns that started or stopped accepting nulls, and enum values that appeared or disappeared. It exists for the failure mode where nothing errors — the pipeline keeps running, the numbers quietly go wrong, and the change that caused it happened three weeks earlier upstream.

Used by

• Data engineers
• Database administrators
• Analytics engineers
• API developers
• Data cleaning teams
• Analysts

Recommended workflow

1. 1

   Turn findings into a contract

   Assert the schema you expect and test future samples against it.

   Open tool
2. 2

   Map old fields to new

   Build the mapping once the drift is understood.

   Open tool
3. 3

   Document the current schema

   Generate a data dictionary from the new shape.

   Open tool

Frequently asked questions

What is schema drift?

Any change to the structure of incoming data — new or removed fields, changed types, changed nullability, new categorical values. It is a problem specifically because it usually does not raise an error; the pipeline keeps running on data that no longer means what it did.

Why is a type change more dangerous than a removed field?

A removed field usually breaks something loudly and gets fixed. A type change from integer to string often parses fine and silently makes sums and averages wrong, which can go unnoticed for months.

How large should my samples be?

Large enough that rare values appear. Nullability and enum drift are both properties of the value distribution, so a small sample will miss them and report false stability.

Can it compare JSON against CSV?

Both formats are profiled into the same internal schema shape, so a comparison will run — but expect noise from structural differences like nesting, which CSV cannot represent.