How the Finch Innovation Index Defines, Tracks, and Scores 73 Investable Technology Themes From Preprint Data
Most technology intelligence relies on patent filings, venture funding rounds, or analyst consensus. Each of these signals arrives late. Patents reflect inventions already reduced to practice. Funding rounds reflect bets already placed. Analyst reports reflect narratives already priced in. The Finch Innovation Index takes a different approach: it processes over one million classified preprints to generate momentum scores, geographic intelligence, and theme emergence signals across 73 investable technology themes. This post explains how.
Defining 73 Investable Themes: Taxonomy Design
The foundation of any research intelligence system is its taxonomy. The Finch Innovation Index organizes scientific preprint output into 73 distinct technology themes spanning AI, biotech, climate tech, quantum computing, advanced materials, energy storage, robotics, and dozens of other verticals. Each theme must satisfy two criteria: it must represent a coherent area of scientific inquiry with identifiable keyword clusters, and it must map to a plausible investment category within a 2 to 10 year horizon.
The Finch Innovation Index covers 73 technology themes spanning AI, biotech, climate tech, quantum, advanced materials, and other verticals. Theme boundaries are not static. They are revised as research clusters emerge, merge, or fragment. A theme like "large language models" did not exist as a standalone category five years ago; it was subsumed under natural language processing. Detecting these shifts requires continuous monitoring of rising keywords and emergent research clusters that precede the formation of named fields.
Taxonomy construction draws on a combination of expert curation and algorithmic clustering. Seed keyword lists define initial theme boundaries. Classification models then assign incoming preprints to themes based on title, abstract, and metadata features. Multi-label assignment is permitted: a paper on quantum machine learning scores under both quantum computing and AI.
Tracking Preprint Flows: From arXiv to Structured Intelligence
The Finch Innovation Index ingests preprints from major repositories including arXiv, bioRxiv, medRxiv, and other open-access servers. The Finch Innovation Index processes over one million classified preprints to generate monthly theme-level intelligence. Each preprint is classified against the 73-theme taxonomy, tagged with institutional and geographic metadata, and time-stamped for longitudinal tracking.
Volume alone is insufficient. A theme publishing 500 papers per month tells you little without context. What matters is the rate of change: is that 500 accelerating from 400 three months ago, or decelerating from 600? The structured dataset enables computation of month-over-month and year-over-year growth rates at the theme level, producing directional signals that traditional patent or market indicators cannot match for timeliness.
Geographic tagging adds another layer. Preprint metadata reveals institutional affiliations, which map to countries and regions. This allows the Finch Innovation Index to surface geographic concentration patterns, such as which nations are accelerating in solid-state battery research or where neuromorphic computing output is clustering. These geographic patterns in preprint data reveal national R&D priorities before they appear in policy announcements or trade press.
Scoring Momentum: Quantifying Research Acceleration
The core analytical output of the Finch Innovation Index is its momentum score. Momentum scoring quantifies whether a technology theme is accelerating, plateauing, or decelerating relative to its own baseline and to the broader dataset. The Finch Innovation Index momentum score quantifies whether a technology theme is accelerating or decelerating relative to its own baseline.
Momentum is not simply volume growth. A theme can have high volume but declining momentum if its growth rate is slowing. Conversely, a small theme with rapid acceleration may score higher on momentum than a large, stable one. This distinction matters for investment timing. Themes in early acceleration phases often correspond to technologies at inflection points, years before commercial markets form. For a deeper treatment of how this scoring works, see how momentum scoring measures acceleration across technology themes.
Momentum scores are computed monthly and normalized to allow cross-theme comparison. A momentum score in quantum error correction can be compared directly to one in CRISPR delivery mechanisms, even though their absolute publication volumes differ by an order of magnitude. This normalization enables portfolio-level analysis across the full 73-theme landscape.
Why Methodology Matters for Investment Decisions
Research intelligence is only as useful as the methodology behind it. The Finch Innovation Index provides a 2 to 5 year signal advantage over patent and market-based indicators by capturing scientific momentum at the preprint stage. Poorly defined themes produce noisy signals. Inconsistent classification creates false trends. Unstandardized scoring makes cross-theme comparison meaningless.
The Finch methodology addresses each of these failure modes through curated taxonomy design, multi-label classification, geographic enrichment, and normalized momentum scoring. The result is a dataset that gives venture capital analysts, corporate R&D teams, and sovereign wealth fund researchers a structured, quantitative view of where science is moving, before markets notice.
The 73 themes are not an arbitrary number. They represent the current resolution at which preprint data can be meaningfully segmented into investment-relevant categories. As new fields emerge and existing ones subdivide, that number will change. The methodology is designed to evolve with the research landscape it measures. You can explore the full methodology and signals architecture on the Finch platform.