# Schema Drift

> When the columns of new data change and no one says so.

**Field:** Data Engineering  
**Category:** Governance & Quality  
**Also known as:** Schema Change, Upstream Schema Change

## Think of it like

Someone moves the chairs while you sleep. You walk in the dark and bump into one.

## Definition

Schema drift is when the shape of incoming data changes without warning. A column is added, renamed or dropped, or its type changes. The source team often makes the change for a good reason and does not know who reads the data. Downstream, a pipeline may crash. Worse, it may keep running and fill a column with blanks.

## Example

The app team renames price to unit_price. The nightly load still runs, but the price column in the warehouse is now empty.

## Why it matters

It is one of the most common causes of broken pipelines. The silent kind can go unnoticed for weeks.

## Common mistake

Letting a loader add and drop columns on its own, with no alert. The pipeline survives, but the reports quietly go wrong.

## Three principles

- **Unannounced change.** Added, renamed, dropped or retyped.
- **Loud or silent.** A crash, or blanks that go unseen.
- **Catch it early.** Contracts and monitors spot it.

## Related terms

- [Data Contract](https://learn.benibix.com/term/data/data-contract)
- [Data Observability](https://learn.benibix.com/term/data/data-observability)
- [Schema Registry](https://learn.benibix.com/term/data/schema-registry)
- [Database Schema](https://learn.benibix.com/term/development/schema)
- [Schema Evolution](https://learn.benibix.com/term/data/schema-evolution)

## Told apart from

- [Schema Evolution vs Schema Drift](https://learn.benibix.com/compare/schema-evolution-vs-schema-drift): People think schema drift and schema evolution are the same change. Evolution is planned: a team adds a field on purpose and checks that readers still work. Drift is a surprise: a source changes its shape and nobody downstream was told. The first is a habit to build. The second is a problem to catch.

![Top: a pipeline with a gate between load and transform, where checks for not null, unique and price above zero let good rows through and stop bad ones. Bottom: monitors watch a table over a week for row count, freshness and schema, and raise an alert when Thursday's row count drops.](https://learn.benibix.com/diagrams/quality-checks-vs-observability.svg)

---

Source: https://learn.benibix.com/term/data/schema-drift
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
