Legal

Data Sources & Methodology

Transparency is a core principle of Axiom Signals. This page describes — at a high level and without exposing proprietary details — where our data comes from and how our analytical outputs are constructed.

Effective: July 12, 2026Last updated: July 12, 2026

Data Sources

Axiom Signals combines data from a rotating set of public and licensed providers, including:

  • Quotes and reference data — Finnhub and other equity data APIs.
  • Macro / economic data — the Federal Reserve Economic Data (FRED) service.
  • News — public news feeds surfaced through our providers.
  • Analytics infrastructure — Supabase (database, authentication) and Cloud Services.
  • AI generation — AI Gateways and underlying large-language-model providers.

The set of providers can change as we improve the platform. We do not currently subscribe to premium, exchange-direct real-time market data feeds; where displayed quotes are delayed, we label them as such.

Freshness & Delay

  • Most equity quotes displayed on the Service are delayed by at least the interval permitted by our providers.
  • Data may be cached for short intervals to reduce load on providers and to protect against transient failures.
  • Timestamps and "as of" labels reflect the best-effort freshness of the underlying data at the moment of display.
  • Fundamentals, macro, and earnings data update on their own natural cadences — often daily, weekly, or monthly.

Source Data vs. Derived Analytics vs. AI Explanation

Every value shown on Axiom Signals falls into one of three categories:

  • Source data — quotes, fundamentals, macro series, and news pulled from providers.
  • Derived analytics — indicators, scores, regime classifications, and evidence tags computed from source data by our own models.
  • AI-generated explanation — natural-language commentary generated by a large language model to describe the derived analytics.

Distinguishing the three helps you evaluate what to trust and what to verify.

Signal Methodology

Axiom continuously evaluates a defined universe of stocks and ETFs. A signal is published when a combination of technical structure, momentum, relative strength, volume, volatility, and market-regime alignment meets a threshold. Signals are tiered by the strength of the underlying evidence and are archived when published.

Evidence Architecture

Each signal and market-outlook view is decomposed into:

  • Supporting evidence — the modules that agree with the thesis.
  • Conflicting evidence — the modules that disagree.
  • Missing data — modules that could not evaluate due to unavailable feeds.
  • Regime alignment — whether the thesis is consistent with the current market regime.

Confidence

Confidence is a model-derived label ("low," "medium," "medium-high," "high") that summarizes the balance of supporting versus conflicting evidence and the coverage of contributing modules. It is not a probability of profit and is not calibrated as such unless explicitly stated.

Track Record Methodology

When a signal is published, it is written to an append-only archive. Its forward performance is then evaluated at fixed time horizons (for example 15m, 1h, 4h, 1d) using publicly observable price data. The Track Record page reports:

  • Number of completed outcomes per horizon.
  • Average forward return per horizon.
  • Performance relative to SPY over the same window.
  • Sample size — always shown alongside every statistic.

Track Record measures the analytical performance of published signals. It does not reflect transaction costs, taxes, spreads, slippage, liquidity, or the returns of a real portfolio.

Limitations

  • Methodology evolves as the platform improves; historical outputs may not exactly match today's methodology.
  • Provider outages can cause temporarily missing or degraded data.
  • Sample sizes for niche regimes or short horizons can be small — read Track Record numbers accordingly.
  • All analytical outputs are informational. See Disclaimers & Risk Disclosure.