ROCKDATAX
Methodology

Evidence before ranking.

RockDatax is designed around a simple principle: a score is only useful when users can understand what data created it, how current that data is, and how confident the system is in the result.

Collect and normalize

News, official company disclosures, regulation, financial statements, market context and structured company data are normalized into a common event and company schema.

Group repeated events

Coverage of the same underlying event is clustered. Repeated stories receive diminishing weight so syndication does not artificially dominate the analytical result.

Map business impact

Events are connected to impact channels such as revenue, costs, capacity, financing, governance, sector exposure and regulation. Time horizon is stored separately.

Build multi-factor context

Member scoring combines event intelligence with official disclosures, regulation, fundamentals, technical context, commentary, macro/sector context and risk/liquidity quality.

Calculate confidence

Data completeness, source verification, freshness and cross-source consistency contribute to a separate confidence measure. Score and confidence are deliberately not the same thing.

Explain the result

Rankings are accompanied by evidence, component scores, risk context and model inputs so the result can be inspected rather than accepted as a black box.

Member score architecture

Seven analytical layers.

20%News & EventsDirection, relevance, novelty, source quality and event repetition.
15%Official DisclosuresContracts, financing, investment, governance, risks and company events.
10%RegulationSPK and other relevant regulatory changes mapped to company exposure.
25%FundamentalsProfitability, growth, leverage, cash flow, efficiency and valuation context.
15%Technical ContextTrend, momentum, volume participation, volatility and relative structure.
10%Commentary / Sector / MacroResearch commentary and broader business context, clearly separated from verified facts.
5%Risk & LiquidityLiquidity, volatility, data gaps and risk flags can reduce the effective score.
SeparateConfidenceSource verification, data completeness and freshness are reported separately.
Weights are a starting framework, not a promise of predictive performance. In production, factor weights should be validated through out-of-sample testing, sector-level calibration and ongoing model monitoring. RockDatax should preserve a full calculation trail for every published score.