Skip to main content

How AI matching works

Otnox uses AI to score every tender on a scale of 0 to 100 based on how well it matches your company's profile, past activity, and bidding patterns. This saves hours of manual screening by surfacing the best opportunities first.

The scoring pipeline​

When a new tender is ingested from a procurement portal, it passes through several stages:

  1. Text extraction — the tender title, description, and metadata are parsed and normalized
  2. CPV matching — the tender's CPV codes are compared against your company profile codes, with exact matches weighted heavily and parent/sibling codes contributing smaller boosts
  3. Semantic analysis — the tender description is analyzed for keyword and concept overlap with your company description and historical bids
  4. Value fit — the estimated contract value is compared to your company's typical bidding range
  5. Buyer analysis — if you have won contracts from this buyer before, the score increases
  6. Final scoring — all signals are combined into a single 0-100 score

What makes scores accurate​

The more information Otnox has about your company, the better the scores. The three most impactful inputs are:

  • CPV codes in your company profile (immediate effect)
  • Company description with specific keywords about your services (immediate effect)
  • Project outcomes — marking tenders as won or lost in your pipeline teaches the model your actual win patterns (improves over time)
tip

The AI engine improves as you use it. Make a habit of marking tender outcomes in your project pipeline. After 20-30 marked outcomes, you will notice a measurable improvement in score accuracy.

Per-company personalization​

Scores are calculated per company. If you manage multiple companies in Otnox, each one receives independently personalized scores based on its own profile and history.

Score refresh​

Scores are recalculated whenever your company profile changes or when tender details are updated by the procurement authority.