Methodology
stockthemes.ai methodology
Purpose and scope
The goal of this site is to break down the themes that are moving and have moved the market over time. Each theme has been hand-crafted by an institutional investment team that has done diligence to determine the right stocks and proper weightings for each theme. We've found Claude and ChatGPT struggle to map public companies into specific market maps. In the text below we explain what gets included, how weights are represented, how returns are computed, and how often data can change
stockthemes.ai is a thematic intelligence product, not an index fund prospectus, portfolio recommendation, or execution venue. Theme baskets are research groupings of public equities tied to a specific narrative.
How themes are constructed
Themes are manually curated from public companies based on documented business exposure to a specific theme. Inputs include filings, earnings commentary, segment disclosures, product roadmaps, and recurring initiative-level signals.
A company is included when there is durable, evidence-based exposure to the theme narrative. Companies can be removed when exposure becomes immaterial, stale, or contradicted by new disclosures.
Groups are umbrella categories containing related themes. Group-level statistics are derived from the underlying themes and their constituents.
How weights are assigned
Theme detail pages include constituent weights used for exposure context and weighted aggregations. Weights are computed from the current theme constituent set and normalize to 100% within each theme basket.
Weights are representation weights for analytical comparison across names within a theme. They are not trade instructions and are not intended to replicate a live investable product.
How performance is calculated
Performance shown across the site is derived from the underlying market data pipeline and normalized to common baselines for comparability. Theme and group pages may show multiple aggregation views (for example average, median, or weighted-average) depending on the endpoint and chart.
Return windows such as 1D, 10D, MTD, YTD, and period-specific ranges are calculated from the corresponding time-series snapshots used by the ETL process. Historical values can be refreshed if source corrections or late-arriving updates occur.
How often baskets are updated
Constituent membership and metadata are reviewed on an ongoing basis and can update whenever new information materially changes thematic exposure. Market-driven snapshot files refresh on a recurring ETL cadence, including intraday updates during market hours and broader pipeline refresh cycles.
Data can be delayed and may be revised. The most recent build timestamp is published in manifest fields such as as_of and optional build identifiers.
Limitations and caveats
Thematic classification requires judgment. Reasonable analysts may disagree on edge cases, especially for diversified businesses with mixed segment exposure.
Nothing on stockthemes.ai is investment advice. Use this methodology together with primary filings, earnings materials, and your own risk process before making decisions.