Use only the data available at the time.

Mostly Right separates when a value applies from when the source published it. Later revisions remain additional records, so a historical cutoff never sees a correction released afterward.

Read the docs

source-history.json

 [
  {
    "indicator": "cpi",
    "period": "2025-01",
    "value": 4.0,
    "knowledge_time": "2025-02-12T13:30:00Z",
    "source": "bls"
  },
  {
    "indicator": "cpi",
    "period": "2025-01",
    "value": 3.9,
    "knowledge_time": "2025-03-12T13:30:00Z",
    "source": "bls"
  }
]

Know what happened and what was knowable.

Event time identifies the period or event. Knowledge time identifies availability. Revision history records what the source changed later.

Separate two kinds of time

Event time states what a value describes. Knowledge time states when the source made that value available.

Keep every revision

A corrected value becomes another published record instead of silently replacing the value available earlier.

Keep the source with the value

Source identity stays beside the value, knowledge time, and revision used in your analysis.

Every time field answers one specific question.

The returned fields keep the event, availability, source, and revision separate so analysis can select the record eligible at a cutoff.

period: The month, quarter, day, or event described by the value.
knowledge_time: The UTC instant when that exact value first became available from the named source.
source: The provider and delivery path that produced the row.
multiple rows: Later revisions remain separate rows for the same period instead of overwriting the earlier print.

Set a cutoff. Select eligible records. Keep the evidence.

Mostly Right does not invent availability timestamps a source never exposed. Reconstructed or unknown timing remains labeled accordingly.

01

Set the historical cutoff

Choose the instant your research, model, or decision is allowed to know.
02

Select eligible records

Use knowledge_time to include only values and revisions published by that cutoff.
03

Keep the evidence

Retain the selected source, knowledge time, and revision with the reproduced result.

Reproduce a result with only the data available then.

Get started See normalized schemas