Today, Theia Insights becomes Onthos. Theia remains in the name of our industry classification system, while Onthos reflects the expanding scope of our work, from mapping the economy to helping firms map their own knowledge.
When we started the company, we chose the name ‘Theia’ after the Greek goddess of sight, in the belief that clear vision leads to sound decisions. At its core, finance is about allocating resources. Doing that well requires seeing the whole: what industries and themes exist and how companies connect to them. In a fast changing and interconnected economy, that whole picture is difficult to assemble, so we set out to build it.
Over the past four and a half years, we have built a suite of market intelligence products: the Theia Insights Industry Classification, Thematic Factor Models and Concept2Universe, which turns ideas into a ranked list of companies. Underneath them all is a shared foundation: a living ontology. An ontology organises information into concepts and categories, and defines how they relate. It enables us to distil vast amounts of information about financial markets into a living map of the economy.
Working with our clients, we increasingly see the same need within their firms. They hold a wealth of information, but much of it is fragmented and disconnected. Bringing it together requires a shared structure, so knowledge from different sources can be connected, queried efficiently and traced back to its source.
So we are making that technology available as Onthos Infrastructure. It builds a living ontology from a firm’s information and distils it into connected facts about companies, activities, themes, risks and other concepts that matter to the firm. Each fact retains a link to its source, allowing people, applications and AI agents to query that knowledge and see the whole picture without rereading the material for every question.
We conducted a study comparing the Onthos approach with RAG (retrieval augmented generation) and with supplying source text directly to an LLM. On company screening and research questions, Onthos achieved F1 scores, a measure combining precision and recall, of 0.90–0.98 across five models, compared with 0.43–0.53 for RAG retrieving eight passages, at a per-question cost comparable to RAG and a small fraction of supplying the documents directly. See the white paper for the methodology and results.
As our work expands, we are choosing a name that reflects our foundation. Onthos is based on ὄν, ὄντος, meaning ‘that which is’, the root of ‘ontology’, while the theta (θ) carries a piece of Theia (Θεία).
To our clients, investors and everyone who has helped us get here: thank you. We look forward to building the next chapter with you. We carry the same belief we started with: good decisions come from seeing clearly, from seeing that which is.