Reeco HorizonHorizon-Scanning AI
Frameworks

Three frameworks.
One continuous corpus.

Every Reeco Horizon framework draws on the same forty-one-million-document corpus, the same event graph, the same calibration discipline. They differ in what they forecast and who they are built for — policy, project pipeline, supply-chain carbon — but the underlying machinery and the underlying commitments are shared.

Framework 01
Horizon Policy
Probabilistic forecasts on upcoming policy instruments across 47 jurisdictions — connection-queue reforms, contracts-for-difference design, planning rule changes, fast-track regimes, consenting timelines. Calibrated quarterly, published with provenance.
PolicyESGForecasting

Open framework →

Framework 02
Horizon Pipeline
A project pipeline graph for renewables — every active project linked to its developer, financier, equipment suppliers and grid constraint. Per-project gate-by-gate connection probability, calibrated against NESO and equivalent outcomes.
Project FinanceDevelopersOEM

Open framework →

Framework 03
Horizon Scope 3
Continuous, activity-based Scope 3 inference for listed companies. Normalised across reporting taxonomies, refreshed monthly, shipped with calibrated uncertainty intervals. Built for sustainability-linked instruments that need numbers tough enough for audit.
CarbonDisclosureListed Equities

Open framework →

What is shared

The commitments behind every framework.

Provenance, every metric
Every output links back to its source documents and the parse that produced the structured fact. Click-through evidence is the default, not the upgrade.
Calibrated uncertainty
Probabilistic outputs ship with 90% intervals and a documented calibration class. Wide intervals on thinly-evidenced classes are themselves information.
Counterfactuals
Every framework output answers "what would this be under three reasonable alternative assumptions?" Sensitivities are baked in, not bolted on.
Public calibration scoring
Every quarter we score our closed forecasts and publish Brier, log-loss, reliability diagrams. Calibration drift is flagged within one cycle.

Read the approach in full → Request a demo