# Selecting Quality Data Streams
Source: https://docs.chain.link/data-streams/selecting-data-streams

> For the complete documentation index, see [llms.txt](/llms.txt).

When you design your applications, consider the quality of the data that you use. Ultimately you are responsible for identifying and assessing the accuracy, availability, and quality of data that you choose to consume via the Chainlink Network. Note that all streams contain some inherent risk. Read the [Data Streams Best Practices](/data-streams/concepts/best-practices) and [Developer Responsibilities](/data-streams/developer-responsibilities) sections when making design decisions.

For a summary of data sourcing models by asset type, see [Data Sources](/data-streams/data-sources).

## Single Source Data Providers

Some Chainlink Data Streams rely on a single underlying data source or a single data provider. This can be necessary when only one authoritative source exists onchain or offchain for a specific type of data, such as a protocol-published NAV, reserve value, redemption rate, exchange rate, or other custom metric.

Single-source streams have a different risk profile than market price streams that aggregate observations across multiple data sources and node operators. If the underlying source or provider publishes incorrect data, omits data, changes its methodology, experiences an outage, is delayed, or is compromised, the stream may reflect that issue. Any such error, omission, delay, or provider failure could negatively affect your application and its users.

Before using a single-source stream, determine whether it is suitable for your use case and whether your application can tolerate the provider/source risk. This is especially important when the stream is used to secure value, determine collateral value, trigger liquidations, support minting or redemption, calculate NAV, or automate other protocol actions.

Users of single-source streams MUST implement additional safeguards such as value bounds, caps, circuit breakers, freshness checks, fallback behavior, monitoring against independent references where available, and manual pause or kill-switch controls. Single-source streams can be classified as High or Very High Market Pricing Risk depending on the asset, methodology, provider dependency, and intended use case.

If you are using or considering a single-source stream in production, [contact the Chainlink Labs team](https://chain.link/contact?ref_id=DataStreams) to discuss the stream's methodology, provider dependency, and appropriate application-level mitigations.

> **NOTE**
>
> For important updates regarding the use of Chainlink Data Streams, subscribe to the [Chainlink
> Changelog](https://docs.chain.link/changelog?product=Data+Streams).