Skip to main content
Cloud Delivery is in limited availability. To request access, reach out to your sales or account manager.

Overview

Cloud Delivery publishes Coin Metrics data as Apache Parquet files in Amazon S3, refreshed daily, and exposes the same data as tables in Snowflake. Rather than paging through an API, you can read whole datasets directly from object storage or query them directly. Quantitative researchers, data engineers, and analytics teams use it to load multi-year histories of trades, candles, quotes, and order books into their own environment or perform analytical queries using Snowflake.

At a Glance

Ten data types are delivered to either target: trades, candles, quotes, order book snapshots, order book updates, reference rates, and four reference data catalogs. Every one is refreshed daily and covers the previous UTC day. The two targets are provisioned separately, so you can take one, the other, or both.
Delivery targetHow you reach the dataFormatPartitioningAccess modelYou pay for
Amazon S3Read objects from a Coin Metrics bucket using your own AWS account and toolingApache Parquet, Snappy compressedHive-style key=value prefixes by market type, exchange, and dateCross-account IAM, read-onlyS3 requests and data transfer, billed to your AWS account
SnowflakeMount a read-only database in your own Snowflake account and query it in SQLTables with native Snowflake column typesPartition columns such as market_type, exchange, and dtSecure Data Sharing, or a private listing for cross-region deliveryWarehouse compute for your queries

When to use Cloud Delivery

Cloud Delivery and the API serve different ways to access our data. The API is the right tool for interactive queries, narrow slices, and anything in real time. Cloud Delivery is the right tool when the volume of data makes repeated API calls impractical.

Data Types

Ten data types are published. Each maps to one prefix in S3 and one table in Snowflake. Field definitions live on the existing data type pages, since the columns delivered here are the same ones the API returns.
Data typeS3 prefixSnowflake tableField reference
Market tradesdata-type=market-trades/market_tradesMarket Trades
Market candlesdata-type=market-candles/market_candlesMarket Candles
Market quotesdata-type=market-quotes/market_quotesMarket Quotes
Order book snapshotsdata-type=market-orderbooks/…/dataset=snapshots/market_orderbooks_snapshotsMarket Order Books
Order book updatesdata-type=market-orderbooks/…/dataset=updates/market_orderbooks_updatesMarket Order Books
Reference ratesdata-type=reference-rates/reference_ratesReference Rate
Reference data: marketsdata-type=reference-data/dataset=markets/reference_data_marketsMarket Reference Data
Reference data: assetsdata-type=reference-data/dataset=assets/reference_data_assetsSee below
Reference data: exchangesdata-type=reference-data/dataset=exchanges/reference_data_exchangesSee below
Reference data: pairsdata-type=reference-data/dataset=pairs/reference_data_pairsSee below

Schema differences

Two data types carry fewer fields than the API returns. Everything else matches the linked pages field for field, on both delivery targets. These fields remain available over the API if you need them.

Reference data schemas

The assets, exchanges, and pairs catalogs each carry two columns, matching the /reference-data/assets, /reference-data/exchanges, and /reference-data/pairs endpoints. The markets catalog is much wider. See Market Reference Data for its fields.

Coverage

Cloud Delivery carries a curated subset of the Coin Metrics market data universe rather than every market on every venue. The scheduled datasets cover the following. Two of these are narrower than the corresponding product documented elsewhere in this knowledge base. Reference rates are delivered at two frequencies against four quote currencies, where the Reference Rate product covers more of both. Market quotes and order books cover spot and futures only, with no options. For the market-level universe behind each dataset, see the coverage tool. If a market, frequency, or quote currency you need is not in the delivered set, your account manager can arrange a backfill, and the API carries the full set in the meantime.
Reference data has no history. The four reference-data catalogs are current snapshots that are overwritten each day. They carry no date partition, so you cannot reconstruct the catalog as it stood on a past date. Retain your own copies if you need point-in-time catalog history.

Update schedule

Every dataset runs once a day and covers the previous UTC day. Jobs start on a stagger to spread load. A file’s dt partition is the UTC date of the data it contains, not the date the job ran. Data for dt=2026-07-28 is produced by the run that begins early on 2026-07-29. Larger datasets take longer to complete, so allow for processing time after the preceding start times before expecting a full day to be present.

Getting access

Access is provisioned per customer and per delivery target. To get started, reach out to your sales or account manager and tell them which target you want: Amazon S3, Snowflake, or both.
  1. Confirm your entitlements and the data types you need.
  2. Provide the details for your target. For S3 this is your AWS account ID and, preferably, an IAM role ARN. For Snowflake this is your Snowflake organization and account identifier.
  3. Coin Metrics grants read access and confirms the bucket, prefixes, or share name you have been given.
  4. Validate the connection using the examples on the target page below.

Choose your delivery target

Amazon S3

Snowflake