How Narwhal keeps official data current

An API request should not have to wait for a government website. Narwhal checks official sources separately, validates each publication, stores the accepted result, and serves that stored data through REST and MCP.

Narwhal checking an official data release before it reaches the API

Your request reads stored data

When you request CPI, GDP, labour, or trade data, Narwhal reads the latest accepted publication from its database. It does not call the official publisher while your request is waiting.

This separates customer response time from the work of retrieving and checking a new release. A slow source website can delay a source check without making every customer request wait on that website.

A release has to pass as a whole

New data becomes public only after the complete release passes validation. The update path is deliberately simple:

  • A scheduled worker checks the official source.
  • Narwhal retrieves the release and checks its identity, period, required fields, and completeness.
  • The normalized observations and their source evidence are stored together.
  • The complete release becomes the current customer version only after every required check passes.

A failed check does not erase good data

Official sites time out. Files can arrive incomplete. Formats change. If a candidate release fails validation, Narwhal keeps the last accepted publication available. It does not replace values with zero or combine part of a failed release with the previous one.

If Narwhal has no publication it can safely serve, the API returns a documented 503 error instead of inventing a result.

Read the dates literally

The period field tells you which month, quarter, or year the observation covers. The released_on field is the official publication date supplied for that release.

The public beta does not expose a Narwhal ingestion timestamp or promise a fixed delay from source publication to API publication. These dates describe the economic observation and its official release, not the speed of your API call.

Revisions replace earlier values

Economic agencies routinely revise published figures. When a revised official release passes the same checks, it can replace the value previously stored for that period. Applications should treat official economic data as published data that can be revised, not as an immutable event log.