Macro nowcasting signal library built on physical-world feeds: 265 daily-updated signals across 12 countries spanning electricity consumption, rail carloadings, nighttime lights imagery, air-quality sensor readings, AIS vessel traffic, satellite-derived parking-lot occupancy, subway turnstile counts, soil-disturbance detection, building permits, spot freight rates, job-posting velocity and transit volume — each aligned to a corresponding official statistical release. The genuinely proprietary byproduct is the alignment layer behind it: a signal-to-release mapping table pairing every daily signal observation with subsequent official printings and their successive revisions, plus the divergence-detection output (per-statistic credibility score, statistical-confidence rating on each divergence, and a revision-tracking record of whether the agency later revised toward the signal), and the discovery engine's backtested signal-composite library for nowcasting employment, inflation, manufacturing and output releases.
Equities, fixed_income, commodities, currencies, derivatives in US, UK, Europe, Japan, China/HK. Related tickers CAT, UNP, NSC, FDX, WMT, TGT, XLNX, SPY.
Tabular daily time series (one series per signal per country) plus satellite imagery rasters behind derived products, structured Text for the divergence and credibility annotations, and API-delivered composites
265 signals x 12 countries x daily updates = roughly 96,700 new observations per year, or about 1.1 million over a decade of history. Add the alignment table: for every paired official release, one print plus each revision — that release-and-revision panel is the compact, high-value object.
Labels that come free with the mechanic: each official release plus its full revision history as a target variable against daily signal observations; binary and scored divergence labels (did the signal disagree with the print, at what statistical confidence); a per-statistic credibility score treating later revisions toward the signal as the ground-truth verdict; realised lead-time labels per signal (the claimed 3-5 weeks to the manufacturing survey, 2-4 weeks to retail sales); market-verdict labels recording whether price action validated the divergence; and a large library of backtested feature combinations with their out-of-sample performance, which is itself a labelled search-result corpus for macro feature discovery.
Alt Signal Intelligence — Macro nowcasting signal library built on physical-world feeds: 265 daily-updated signals across 12 countries spanning electricity consumption, rail carloadings, nighttime lights imagery, air-quality sensor readings, AIS vessel traffic, satellite-derived parking-lot occupancy, subway turnstile counts, soil-disturbance detection, building permits, spot freight rates, job-posting velocity and transit volume — each aligned to a corresponding official statistical release. The genuinely proprietary byproduct is the alignment layer behind it: a signal-to-release mapping tab…
Not yet on file: Sample · Data dictionary · Data frequency · Reporting lag · Point-in-time · License