When people hear that an AI system matches tenders to companies, they often imagine simple keyword matching. The reality is significantly more sophisticated.
The problem with keyword matching
Keyword matching fails because tender language and company descriptions rarely overlap precisely. "Civil engineering services for highway rehabilitation" will not match "road infrastructure maintenance." The words differ. The meaning is the same.
Vector embeddings: turning text into numbers
The foundation of modern AI matching is the vector embedding, a mathematical representation that captures meaning, not just words. Two texts with similar meanings have similar embeddings even if they share no words.
The matching process
When Trinta processes a new tender, it generates an embedding from the title, description, category, and issuer. When a company completes their profile, an embedding is generated from that too. Finding matches is a mathematical similarity search across the tenders your agents collect every day, returned in milliseconds. The approach was built on an archive of more than 450,000 real tenders.
The five-factor scoring system
Semantic similarity alone is not enough. Trinta combines it with four additional factors:
Industry overlap (25%): Do the company's industries match the tender's categories?
Geographic fit (15%): Is the company active in the tender's country?
Company size fit (12%): Does the contract value match company capacity?
Historical signals (8%): Has the company won similar contracts before?
The feedback loop
Matching improves through user feedback. Thumbs up and thumbs down signals improve future matches for similar companies.
Frequently asked questions
01How does AI tender matching actually work?
AI tender matching converts both the tender text and a company's product description into vector embeddings, mathematical representations of meaning, then measures how close they sit. Because the comparison is semantic rather than literal, a notice that describes your product in different words still matches. Scoring then layers on factors beyond similarity, such as geography and contract scale.
02What are vector embeddings in procurement matching?
A vector embedding is a numerical representation of text that places similar meanings close together in mathematical space. In tender matching, both the notice and the supplier's catalogue become embeddings, so the system compares meaning rather than exact wording. This is what allows a differently-phrased specification to match a product it genuinely fits.
03Why does keyword matching miss relevant tenders?
Because buyers describe what they need in their own words, not yours. A hospital tendering for oxygen plant may write medical gas production unit, PSA generator, or a bare technical specification with no product name at all. A keyword list matches only the phrasing you predicted, and floods you with notices that happen to share a word.
Share this article
Reference
Related articles