Customer data glossary

Event stream

By Art Kukharenko · Published · Updated

Browse learn glossary resources

An event stream is an ordered, continuous flow of timestamped records, such as page views, transactions or consent changes, that a platform ingests and processes as they occur.

Part of the Learn hub and the customer data glossary.

What it means in practice

An event is a record that something happened: a page was viewed, a payment cleared, a consent changed. It carries who (an identifier), what (a name), when (a timestamp) and properties (the details). A stream is the ordered flow of such events. Log-based brokers such as Apache Kafka retain events for a configured period, so consumers can read at their own pace and re-read from an earlier point.

Streams are the nervous system of real-time decisioning and of activation that reacts as things happen, in contrast to batch loads that move data in scheduled chunks. Design points: partition by person so each person's events stay in order; expect at-least-once delivery, so duplicates will occur and consumers must be idempotent; handle late and out-of-order events; and put every event through an event contract at the door.

Consent changes belong in the stream. A withdrawal is an event like any other, and processing it promptly is what makes suppression work in near real time.

Keep out what should not be there. Cardholder data in event payloads may bring pipelines into PCI DSS scope, and free-text fields can carry health information nobody meant to collect.

Example (retail banking): app and branch events flow into the stream as event types and tokenised references, not card numbers. A decisioning service reads them and suppresses a promotion if a fraud alert has just fired on the account.

Where it matters

Frequently asked questions

How is an event stream different from a message queue?

A queue usually delivers each message once to a consumer and discards it. A log-based stream retains events so several consumers can read them and replay them.

Do I need Kafka?

Not necessarily. Smaller volumes can use simpler pipelines. Choose by volume, ordering, retention and replay needs.

Is an event stream personal data?

Usually yes, because events link to identifiers. Apply minimisation, retention limits and deletion to it like any profile store.

Back to the customer data glossary or the Learn hub.

Further reading

Published: · Updated:

Publication date recorded for the approved release. Editorial updates do not establish legal approval or a customer result.

Need help?

Ask the PrivyCDP Assistant