Technology
Models that explain themselves
Markets don't reward black boxes that no one can question. We combine transparent scoring rules with machine learning, and keep every output traceable to the data that produced it.
Official sources
Scheduled collectors pull each release as it publishes and record when it arrived.
Normalize
Series are aligned to a common calendar, revisions are tracked, and gaps are flagged instead of filled silently.
Transform
Levels become changes, percentiles, trends and ratios, such as margin debt relative to GDP.
Score and detect
Rules score each indicator; statistical models look for shifts across the whole set.
Human check
Analysts review changes before they reach research, and add readings that data alone can't capture.
Transparent scoring
Every status has a rule you can read
Each indicator on the Macro Monitor has a written rule. For example, the 10-year Treasury yield scores as downside risk at 5% or above and borderline above 4.5%. Margin debt relative to GDP is scored by its percentile since 1997.
Because the rules are explicit, you can see why a status changed and decide whether you agree. Rules are reviewed as market structure changes, and each revision is dated.
Machine learning
Where models add signal
Regime detection
Classifying the current mix of conditions and comparing it with past episodes, such as rate-driven slowdowns and leverage-driven selloffs.
Nowcasting
Estimating monthly and quarterly releases from faster weekly and daily data before the official number arrives.
Positioning analytics
Measuring how stretched futures positioning is against its own history, and how prices have behaved from similar extremes.
Text analysis
Reading central-bank statements and energy-agency reports to track changes in tone and emphasis over time.
Data provenance
Where the live data comes from
Public figures on this site come from official sources and are shown with attribution. Where an input comes from a licensed provider, we publish our assessment rather than the provider's numbers.