C
Docs

Similar Opportunities

Semantic matching to related opportunities and similar awarded contracts

Updated 2026-07-223 min read

Similar Opportunities

Cothon uses semantic embeddings to relate an opportunity to other opportunities and to historically awarded contracts. This gives you two complementary panels on each opportunity: similar opportunities and similar awarded contracts.

How Similarity Works

When an opportunity is saved, Cothon generates a semantic embedding — a numerical representation of its meaning built from its title and description. Similarity searches compare embeddings, so matches are based on what the work is, not just shared keywords.

Note

Because matching is semantic, an opportunity described as "IT professional services" can match one described as "informatics staffing" even though the words differ.

Similar Opportunities Panel

From an opportunity, the similar-opportunities view lists other opportunities ranked by semantic similarity. Each match shows a similarity score, and a minimum-similarity threshold keeps weak matches out of the list.

Typical uses:

  • Pipeline building — found one good fit? See what else looks like it.
  • Pattern spotting — recognize recurring buys from the same kind of requirement.
  • Proposal reuse — identify pursuits where past proposal content may transfer.

Similar Awarded Contracts

The opportunity workspace also surfaces similar awarded contracts — historical awards whose content resembles the current opportunity. For each match you can see the winning vendor and award details from the public record.

Use this panel to answer:

  • What did comparable work actually sell for?
  • Which vendors win this type of contract?
  • Is there a likely incumbent?

Tip

Similar awarded contracts also feed the pricing intelligence shown in the opportunity's Intelligence tab — see Opportunity Intelligence.

Before you save a tender, the Tenders detail view shows a historical-awards panel keyed on the tender's UNSPSC codes. After saving, the opportunity's similar-awarded-contracts panel takes over with semantic matching. The two views complement each other: one is category-based, the other meaning-based.

Notes and Limits

  • Similarity requires an embedding; embeddings are generated when opportunities are saved, and can be refreshed if an opportunity's description changes materially.
  • Match quality depends on how descriptive the opportunity text is — a one-line description gives the matcher little to work with.
  • These panels are for on-screen research; there is no export of similarity results.

Next Steps

Was this page helpful?