Intelligent tender scoring – automatic selection of the best business opportunities

About client

HAMMERmed Medical Polska is one of the market leaders in the distribution of medical devices in Poland. The company actively participates in public procurement procedures, offering a broad portfolio of products to medical facilities across the country.

A leading distributor of medical products and modern treatment methods on the Polish market.

About project

HAMMERmed analyzes a large number of tenders that vary in scope, documentation format, and level of alignment with the company’s product offering.

Previously, the tender evaluation process was largely based on manual document analysis, which resulted in:

• high involvement of sales and operational teams,
• the risk of missing attractive tenders,
• a lack of consistent and repeatable criteria for assessing profitability,
• limited scalability as the number of procedures increased.

The client needed a solution that would automatically evaluate tenders in terms of fit with the product portfolio, identify the most promising opportunities, and genuinely support the decision-making process of business teams.

3Soft designed and implemented an automatic tender scoring system based on its proprietary DocMiner module, supported by AI and LLM mechanisms.

The solution analyzes tender documentation (emails, PDFs, scans), identifies products included in the tender, matches them to HAMMERmed’s product groups, and calculates a numerical score indicating the attractiveness of a given procedure.

The system automatically:

• rejects tenders with low alignment,
• routes the most promising procedures for further analysis (based on defined acceptance thresholds),
• learns from user decisions, increasing the accuracy of assessments in subsequent iterations.

Importantly, the client has access to a web application where they can:

• review scoring results,
• approve proposed product-to-product-group mappings,
• manually correct incorrect assignments.

Each user decision is stored and used to train the model, resulting in:

• automatic correction of future results,
• gradual improvement in scoring accuracy,
• reduced need for manual verification in subsequent system iterations.

As a result, the system provides real support for business decision-making, rather than serving merely as an analytical tool.

The project was delivered in two main phases: conceptual and analytical phase and implementation phase.

Phase I – conceptual and analytical, which included:

• analysis of tender data,
• selection of matching methods (LLM vs. statistical methods),
• design of the solution architecture,
• definition of result structures and acceptance thresholds.

Phase II – implementation (scoring), which included:

• development of a web application,
• automation of tender retrieval from emails,
• integration of DocMiner with the scoring engine,
• training the model based on user decisions,
• deployment of the environment in Azure and preparation of the production system.

The solution was built on the Microsoft Azure cloud and includes:

• a containerized processing environment,
• ETL components for data ingestion and preparation,
• a relational database and a vector database,
• local LLM models for document content analysis,
• integration with the client’s systems (including DMS).

Tender data is retrieved fully automatically from a dedicated email inbox, and the entire process – from document reading to score calculation – is performed without end-user involvement.

The use of our proprietary DocMiner solution enabled:

• conversion of documents into an analyzable format (from PDF files to images),
• analysis of document layout (sections, tables, columns),
• extraction of product-related content using AI and language models,
• storage of results in a structured format (JSON), ready for further scoring processing.

Key results

The implementation allowed HAMMERmed to:

• significantly reduce tender analysis time,
• focus teams on tenders with real business potential,
• reduce the number of decisions made “by intuition”,
• automate and standardize the procedure selection process,
• prepare the process for further scalability as the number of tenders grows.

The system has become a real tool supporting sales and business decisions, rather than just another reporting system.