What is a Company DNA Vector? Representing Target Personas in High-Dimensional Semantic Space
TL;DR Summary"Keyword-only lookups miss key prospects that use different words to describe the same pains. CoSDR uses text-embedding-004 to represent ICPs as high-dimensional semantic DNA vectors, executing low-latency local queries using pgvector 0.7 halfvec HNSW compression in PostgreSQL."
1. Topic Context & Definition
A Company DNA Vector is a mathematical representation of a business's technographic, firmographic, and pain-point characteristics within a high-dimensional vector space.
Moving Beyond Keyword Match Prospecting
Traditional B2B prospect filtering relies on rigid keyword matches. If a company describes its business using slightly different terms, standard search systems miss them. Modern outbound platforms solve this by converting target personas into high-dimensional semantic vectors. This enables highly advanced B2B lead finder techniques that bypass static, decayed datasets like those in our CoSDR vs Apollo comparison.
pgvector halfvec Compression in PostgreSQL
Storing high-dimensional vectors (768 or 1536 dimensions) requires substantial database memory. To optimize queries, CoSDR uses pgvector 0.7 halfvec (16-bit float) compression. This reduces the memory footprint of embeddings by exactly 50%, ensuring HNSW indexes reside in PostgreSQL shared buffers for fast local lookup.
Reciprocal Rank Fusion (RRF) Queries
To combine the benefits of semantic and keyword searches, CoSDR runs concurrent queries. The system merges keyword filters with vector similarity scores, combining results using Reciprocal Rank Fusion (RRF) in under 100 milliseconds to find the highest-affinity target accounts.