# Kappa Architecture

> One stream does all the work. Old data is played again when rules change.

**Field:** Data Engineering  
**Category:** Foundations  
**Also known as:** Kappa

## Think of it like

It is like a video you can rewind. To fix a mistake, you play the tape again from the start.

## Definition

Kappa architecture uses only a stream, with no separate batch path. Every event is kept in a log that can be read again. Live data flows through one stream job. When the logic changes, you replay the old events through the new job. So there is one set of code to write and fix.

## Example

A team changes how it counts active users. It starts the new job at the oldest event in the log and lets it catch up.

## Why it matters

It removes the double code of Lambda. One job means one place for bugs and one place to fix them.

## Common mistake

Not keeping events long enough. If the log only holds seven days, you cannot replay a year.

## Three principles

- **Stream only.** No separate batch path.
- **Keep the log.** Old events can be read again.
- **Replay to fix.** New logic reruns over old events.

## Related terms

- [Lambda Architecture](https://learn.benibix.com/term/data/lambda-architecture)
- [Stream Processing](https://learn.benibix.com/term/data/stream-processing)
- [Event Sourcing](https://learn.benibix.com/term/development/event-sourcing)

## Told apart from

- [Lambda Architecture vs Kappa Architecture](https://learn.benibix.com/compare/lambda-architecture-vs-kappa-architecture): People think Kappa is just Lambda with a part cut off, and it is more than that. Lambda runs the same logic twice, once in a nightly batch and once live, and joins the two. Kappa runs it once, in a stream. When the logic changes, Kappa replays its log from the start instead of keeping a batch path.

## Taught in

- [Lambda and Kappa Architecture](https://learn.benibix.com/learn/data-foundations/lambda-and-kappa-architecture): Data Engineering Basics, 6/11

![Left: Lambda, where new data splits into a batch layer and a speed layer that meet again in a serving layer. Right: Kappa, where all data flows through one stream layer into the serving layer, and old data is replayed through the same stream.](https://learn.benibix.com/diagrams/lambda-vs-kappa.svg)

---

Source: https://learn.benibix.com/term/data/kappa-architecture
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: 2026-11-20
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
