How the numbers are built.
Our operating rule is straightforward: structured data provides the facts, server-side analytics provides the calculations, and AI provides interpretation.
Last updated October 14, 2026Source hierarchy
The public indicator directory contains 26 configured entries: 25 through Federal Reserve Economic Data (FRED) and 1 from the World Bank. FRED distributes data from original publishers including BLS, BEA, Census, EIA, and the Federal Reserve. Directory entries are not a warehouse-wide inventory or proof of complete history. The 17-series revision-sensitive ALFRED refresh rotation is a separate scope; daily Treasury histories and World Bank gold are not part of that vintage rotation. See each indicator’s source link and Data Refresh Status for coverage and freshness.
Where a feed is unavailable or fails validation, the correct state is unavailable. We do not ask an AI model to fill in a missing number.
Public indicator coverage
Inflation: CPI, Core CPI, PCE Price Index, and Core PCE Price Index. Growth and activity: Real GDP, Retail Sales, Industrial Production, Housing Starts, and Building Permits. Labor: Unemployment, Nonfarm Payrolls, and Labor Force Participation. Policy and liquidity: Effective Fed Funds and M2. Rates: 3-month, 2-year, 5-year, 10-year, and 30-year Treasury yields. Additional directory entries cover wages, initial claims, job openings, all-commodities producer prices, WTI spot crude, the broad dollar index, and World Bank monthly gold.
Transformations and derived metrics
Depending on the series, the site calculates level, month-over-month, quarter-over-quarter, year-over-year, annualized quarterly growth, basis-point changes, yield-curve spreads, rolling correlations, z-scores, percentile ranks, OLS regressions, and lead/lag relationships.
Daily Treasury data is normalized to an appropriate common frequency before monthly or yearly transformations are calculated. That avoids treating a one-day move as a month-over-month or year-over-year change.
Revisions and vintages
Economic data can be revised. The warehouse preserves retrieved observations and is designed to support revision-aware research. Revision-sensitive macro series are now being populated with ALFRED real-time vintages. The system stores ALFRED vintages separately from the current snapshot. Revision displays compare the earliest stored vintage with the latest stored vintage for the indicated observation period. The earliest stored vintage is not necessarily the original release; original-release coverage must be verified separately. Retrieval time is not publication time. Historical coverage is seeded automatically on a rotating schedule, so some series may show “building” until their initial backfill completes.
Scenario engine
The Macro Scenario Engine assigns independent 0–100 evidence scores to Soft Landing, Reacceleration, Stagflation, and Recession Risk. The scores are not calibrated probabilities and do not sum to 100. Each score is built from explicit rule-based contributions using verified growth, inflation, labor, policy, housing, yield-curve, and historical-analog inputs.
The current score can include a small adjustment from the 12-month outcomes of the five closest historical analogs. The historical scenario chart excludes that analog adjustment so past observations use the same core rule set. Historical scenario scores currently use latest revised data rather than vintage-pure history.
Research Lab and historical As Of mode
Research Lab supports both current-history studies and a historical As Of mode. In As Of mode, revision-sensitive macro series are reconstructed only from ALFRED vintages available by the selected knowledge date. Non-revision historical market series, such as Treasury yields, are truncated at that same date. If vintage coverage is not available, the study remains unavailable rather than substituting today’s revised history.
Correlation and regression are descriptive statistical tools, not proof of causation. Results depend on the selected sample period, transformation, frequency, revision history, and data quality. Lead/lag relationships can shift across regimes and should be treated as research prompts rather than trading signals.
Freshness and release dates
Economic observations refresh through scheduled ingestion, while the release calendar uses official FRED release-date metadata for major reports in the existing MVP universe. News headlines are a separate discovery layer and are never used as the source of the verified economic values shown in the dashboard.