A forecast earns trust after reality answers.

Which long-horizon programs and partnerships are tracking as intended, and which need to be rethought before the next funding cycle. The calls that resolve over decades and have to defend themselves year after year.

A forecast registry your board can inspect, year over year, with a calibration trend that improves as the program runs.

• ·Held-out backtest at 297 pairs (2020 → 2022/2024/2026)
• ·Brier score per claim with Murphy decomposition
• ·139 verified corrections to historical baselines, each dated and sourced
• ·Rank correlation ρ = 0.906 across the 297 pairs

NIC Global Trends, CSIS, and RAND scenario work. Those disclaim prediction and refresh narrative scenarios every four years. This produces scored, continuously updated forecasts with the audit trail institutional review demands.

Where momentum is real across regions, and where it's local elite consensus that won't generalize. The cross-region read that tells you which signals reinforce each other across the map and which are isolated to one cluster.

A one-page cross-region brief naming the three highest-confidence subject-region intersections for the question on your desk, with the data trail underneath.

• ·9 subjects × 11 regions = 99 surfaces, each on the same 0-to-100 calibration
• ·Sources include IMF, BIS, World Bank, IPCC, IEA, Edelman, WHO, NOAA, ACLED, Stanford HAI, Epoch AI, UNESCO
• ·Positive signals tracked as a first-class surface, not an afterthought
• ·Cross-family ensemble of five forecasters, skew-adjusted aggregation

Verisk Maplecroft and other current-state index families. Those publish current-state only, no trajectory, no Brier. This adds the 30-year forward view and the per-claim accuracy score those don't produce.

The tactical calls that get voted on and have to survive minutes-of-meeting scrutiny. Industry-mix shifts, workforce projection, infrastructure justification, appropriations defense. Reads where the question after the vote is always where did this number come from.

A forecast appendix with a Brier score per claim and a held-out backtest disclosure that survives audit-committee inspection. A scenario range for shock years, a point estimate for calm years.

• ·Held-out 2025 backtest — the model only saw data through 2024
• ·Murphy decomposition on every claim
• ·No retroactive tuning. Model versions tracked separately.
• ·139 verified corrections, dated and source-linked
• ·ForecastBench-schema aligned (Karger et al. 2024)

Eurasia Group, EIU, S&P Global consulting reports for the board-and-commission context. Those publish no held-out backtest and no Brier on their methodology pages. This produces the verify-me artifacts that trust-me advisory cannot.

Industry-mix shifts. Site selection. Workforce projection. Whether to commit to a single sector outlook or hedge across a range. The calibration tells you when the regime is calm enough to act on a point estimate and when the regime is in shock and you need a scenario range.

A forecast appendix that names the regime (calm or shock), gives you a number or a range with that regime's measured uncertainty, and carries a Brier history on the exact claim.

• ·Regime-aware backtest. 2000-2024 expanded economy sweep, 3,575 pairs.
• ·Calm window scored at ±5.5, shock window at ±9.8, separately
• ·Tightest measured slice: Tech · Quantum at ±2.8 on the 2025 hold-out
• ·Widest measured slice: Money · Crypto at ±8.9 on the 2025 hold-out
• ·Cross-family ensemble, weighted by track record not reputation

Oxford Economics (87 economies, economic-only, no Brier) and EIU (5-year horizon). This adds 30-year horizon and cross-domain STEEPE signals into the same calibration so an industry-mix call factors in the forces that bend it, not only its own historical trend.