Methodology

Version 2026.07.2

What the Connection Score means

The Connection Score (0–100) measures how strongly a public instrument connects to the searched company or theme. It blends twelve deterministic components — business-model similarity, revenue relevance, economic dependency, ownership, partnership strength, supplier importance, customer relationship, theme purity, ETF portfolio weight, evidence quality, recency, and substitution risk — with weights that vary by relationship type. Every component is computed from stored, auditable facts; a relationship with weak evidence is hard-capped regardless of how plausible it sounds.

What it does not mean

The Connection Score is not expected return, not a quality or safety rating, not a price target, and not a buy, sell, or hold recommendation. A score of 85 means the relationship to the thesis is strong and well evidenced — the instrument can still perform terribly. Two instruments with the same score can behave in completely different ways, which is why every result also carries a behavior profile (equity growth, credit income, commodity price exposure, and so on).

Relationship categories

Results are organized into ten route categories rather than one blended list: similar public companies; strategic investors and owners; commercial partners and customers; suppliers and infrastructure providers; competitors and substitutes; diversified etfs and exchange-traded products; credit exposure; real-asset and commodity exposure; listed private-market vehicles; foreign and otc listings; disruption and scenario exposure; advanced instruments. Each captures a different mechanism of exposure with different fidelity — a similar public company tracks the business model, a supplier tracks one revenue line, an ETF dilutes everything it holds.

Evidence hierarchy

Every claimed relationship must carry evidence records with source, type, dates, and a quote or summary. Primary sources (regulatory filing, annual report, earnings call, investor presentation, government contract, fund holdings, company website, supplier disclosure) outweigh secondary ones (press release, partnership announcement, news source, reasoned inference). Reasoned inferences are allowed but always labeled as such, score-penalized, and never counted as primary. A map with too little non-inference evidence is not auto-published — it goes to editorial review.

How ETF exposure is evaluated

Funds are scored on theme purity and on the portfolio weight of thesis-relevant holdings, not on holding a single name. In portfolio-exposure analysis a fund position counts toward a thesis only at its thesis-relevant weight (or a conservative default when the weight is unknown) so indirect exposure never inflates the numbers.

Bonds are not stocks; commodities are not companies

Credit routes expose you to coupons and repayment, not growth — a supplier's bond can be money-good while its equity halves. Commodity routes expose you to a price, not to any company's execution; futures-based products add roll mechanics (contango and backwardation) on top. StratSeek labels these behavior differences on every card and refuses to blend them into one ranking.

Freshness and corrections

Every map records data freshness and each relationship a last-verified date; stale evidence triggers review alerts. Users can submit corrections on any page — corrections are reviewed by an editor, and accepted ones are applied with an audit-logged reason and a new map version in the change history.

Conflicts of interest and sponsorship

StratSeek accepts no payment for placement, scoring, or ranking. No issuer, fund sponsor, or broker can pay to appear in, or be removed from, a Strategy Map. If sponsored placements are ever introduced they will be visually separated and labeled as sponsored, and they will never affect Connection Scores.

Limits

Maps are research, not truth: coverage is not guaranteed to be complete, relationships change without notice, and no individualized investment advice is given anywhere on StratSeek. AI assists the research pipeline — proposals are validated against deterministic rules and evidence requirements before anything publishes, and model provenance (provider, model, prompt version) is stored with every map.