HOW COMMON MEASURE WORKS

Context you can check.

Our sources, publication rules, and limits are part of the story.

Start with the original release

We prioritize official aggregate statistics from BLS, BEA, Census, and the Federal Reserve. The publication began with six verified releases from August and September 2026. Automatic explainers use a 30-day freshness window. The date of the economic activity is often earlier than both the agency release date and the Common Measure publication date.

Original explanations use released facts, not copied journalism. We do not bypass paywalls, use agency seals, republish restricted images, or claim agency endorsement. The automatic service reads the official BLS public data API and BEA release pages. These public sources do not require an API key. Census and Federal Reserve items currently remain dated snapshots.

Sources approved for automatic checks

Census housing, trade, and the Fed policy rate are verified launch snapshots. Their automatic extraction is not yet enabled. No secondary reporting sources are approved at launch. Developing stories without primary evidence are held; support for corroboration from two independent reporting organizations is a later addition.

Behind the Story and Econ 101

Behind the Story pairs links to publisher coverage with Common Measure’s independent economic topic guides. Publisher names identify reading context, not partnerships. Where full article access has not been verified, the guide says so and does not claim to summarize the complete argument. Its factual and conceptual explanations cite separate primary sources. Original subscription requirements still apply.

No WSJ, Bloomberg, Financial Times or Economist article-scraping service is enabled. Automated ingestion, reuse and model processing require an appropriate permission or license review for the specific source. A subscription by itself is not treated as permission to republish or automate processing. Curated guide updates are separate from the official-statistics scheduler.

Econ 101 provides definitions, analogies, hypothetical examples, common misreadings and supporting methodology links. Analogies simplify a mechanism. They do not substitute for the underlying definition or constitute a forecast.

Authorship and contribution records

Launch reading companions are labeled AI-generated and not human reviewed. The private writing desk records personal review separately from substantial personal authorship, retains prior versions and requires a contribution note. Editing prose does not automatically validate new claims. An owner-authored label is a self-attested contribution record, not independent certification. Automated publications are not counted as the owner’s authored articles.

What the automation does

The implemented service checks approved feeds and release pages, records source and retrieval dates, identifies the observation period, and extracts a small set of supported statistical sentences. It groups events by report family and period and prevents repeat publication through database identifiers.

It generates an original explanation from AI-authored, fixed templates. There is no runtime AI-model call and no AI API charge in this mode. Labels identify automatically generated articles and never describe them as human reviewed.

Numbers are re-extracted from the statistical sentences with code. Dates, embargo times, units, reference periods, freshness, and conflicts must pass explicit checks. Unknown wording, missing evidence, a changed figure in the same release, or retraction language goes to the owner queue. These checks reduce certain errors; they do not guarantee accuracy.

Source material is treated as untrusted data, not instructions. Only approved HTTPS addresses can be fetched, redirects are checked, response sizes and time are bounded, and external text is rendered as text.

Schedule and current status

A scheduled service is deployed in the owner's Cloudflare account. It wakes every 30 minutes and checks sources hourly by default, with a daily health assessment, a configurable four-article daily cap, and a pause control. The public service status reports whether scheduled checks have actually been observed. A configured schedule alone is not proof of a successful update, and a source check is not a new statistical release.

Statistical integrity

We distinguish monthly and annual changes; percentages and percentage points; current dollars and inflation-adjusted values; adjusted and unadjusted series; and preliminary and revised estimates. Annualized quarterly GDP is not a forecast. Lower positive inflation means prices rise more slowly.

Charts provide values and a data table. Every calculator labels hypothetical assumptions and excludes them from observed data. A central-bank target is not a consumer borrowing rate.

Corrections and updates

Meaningful changes preserve the previous article in a revision history. An official new estimate is labeled as an update. An error correction requires a visible note. A source conflict is flagged for review; the owner can correct or unpublish an article. We do not change timestamps merely to appear fresh.

Existing content remains stored if a source is unavailable. Failed checks and stale observations are visible. Articles retain their original release vintage unless explicitly updated.

Questions and correction requests

No public submission form or email destination is configured yet. We do not display a contact form that cannot deliver. The owner can manage corrections in the protected editorial interface.

Understand why economic data are revised ↗