FRED · BLS · BEAIn development

Macro data as it was known then.

Query economic releases, original values, later revisions, and publication timestamps without leaking today's knowledge into yesterday's backtest.

Join early access Inspect the schema

Free workspace · No card required

DATASETfred.macro_releases

PIPELINE PREVIEW

ILLUSTRATIVE RELEASEUS nonfarm payrolls · monthly change

PAYEMSobservation: 2024-10

INITIAL2024-11-01 12:30 UTC+118K

REVISION 012024-12-06 12:30 UTC+132K+14K

REVISION 022025-01-10 12:30 UTC+129K−3K

AS KNOWN AT2024-11-15T00:00:00Z+118K

DATASETfred.macro_releases

SOURCESALFRED · BLS · BEA

UPDATEPer release

REVISION POLICYAppend-only

01 / BUILT FOR MACRO RESEARCH

Backtest the release.
Not the revised history.

The number published at 08:30 and the number visible six months later are different facts. KnownAt keeps both.

Test release reactions

Join observations to the publication moment instead of an approximate calendar date.

Measure revisions

Compare initial releases, routine revisions, and benchmark updates without silent overwrites.

Avoid look-ahead bias

Recreate the economic information set that a model could actually have seen then.

02 / VINTAGE-AWARE SCHEMA

One observation.
Every published version.

Observation time and information-availability time remain separate. The sample below illustrates the planned normalized model; final fields may change before release.

SCHEMAfred.macro_releases

published_attimestamp

When the release became public and usable.

observation_datedate

Economic period described by the observation.

series_idstr

Stable identifier from the publishing source.

vintage_datedate

Date this version entered the public record.

valuefloat64

Value reported in this specific vintage.

revision_numberint32

Ordered revision index; zero is the initial release.

SAMPLE VINTAGESUTC · illustrative

published_at timestamp	series_id str	vintage str	value float64
2024-11-01 12:30	PAYEMS	initial	118.0
2024-12-06 12:30	PAYEMS	revision_1	132.0
2025-01-10 12:30	PAYEMS	revision_2	129.0
2025-02-07 12:30	PAYEMS	benchmark	126.0

03 / SOURCE COVERAGE

Published statistics in.
Queryable vintages out.

KnownAt preserves source identifiers, release metadata, footnotes, preliminary flags, and each point at which a value became public.

FRED and ALFREDSeries observations, real-time periods, and historical vintage dates.

BLS releasesLabor time series, periods, values, footnotes, and preliminary flags.

BEA statisticsProgrammatic economic datasets plus source metadata and parameters.

Publication provenanceSource, series, release, observation, and availability timestamps.

Official sourcesreleases · series · revisions

KnownAtcapture · version · timestamp

Your modelas-known-at query

04 / POINT-IN-TIME ACCESS

Ask what the model
could know that morning.

The same dataset supports current values, the complete revision chain, and historical information sets bounded by an availability timestamp.

Python

knownat

payrolls = knownat.dataset(
  "fred.macro_releases"
)

rows = payrolls.history(
  series_id="PAYEMS",
  as_known_at="2024-11-15T00:00:00Z"
)

REST JSON Parquet

RETURNED VINTAGE

Only the initial release is visible.

The December and January revisions exist in storage, but they arrived after the requested information boundary and cannot enter the result.

revision_number=0 · value=118.0

05 / AVAILABILITY

Macro vintages are next in the pipeline.

This page shows the planned normalized interface. Join now to follow source coverage, schema decisions, and release progress.

Join early access See founding pricing