How Chainlink Data Feeds work

The essentials

Data Feeds aggregate observations and expose results to applications, commonly through a proxy pointing to an aggregator. This separation can support upgrades without changing the address consumed by an application. It also means that administrative arrangements matter.

How it works

Publication may depend on a deviation threshold or heartbeat interval. A feed is not necessarily a tick-by-tick price stream. Consumers should check the exact asset, network, decimal scale and timestamp, and define what happens when data becomes stale or unsuitable.

What to watch

For example, treating an eight-decimal answer as an eighteen-decimal answer creates a severe valuation error even when observations are accurate. Multiple operators do not eliminate correlated data failures. On a Layer 2, applications may also need to account for sequencer downtime and a grace period before resuming sensitive actions.

Understand the details

A price feed is an interface to a specific reported value, not a universal market price. Integrations must confirm the network, feed address, base and quote assets, decimal scale and update time. Updates may depend on a time interval or a deviation condition, depending on the feed. A transaction can execute between two updates.

Boundaries and common mistakes

Freshness limits should match the application’s risk and feed configuration. Rejecting every small delay can stop a service unnecessarily; accepting arbitrarily old values can misprice collateral. Some L2 integrations also need to consider sequencer availability and recovery behavior. A fallback must have its own validation rules.

The mechanism at a glance

Continue learning

Look up a term in the glossary →