Carrier recommendation system for Trans.eu: intelligent matching powered by data and process modeling.
About the project
Trans.eu is one of Europe's largest freight exchange platforms, connecting freight forwarders with carriers. The platform handles tens of thousands of transactions daily and has continuously expanded its ecosystem of digital services for the transportation industry. Since 2017, fireup.pro has collaborated with Trans.eu as one of several engineering teams contributing to the platform's development, including Team Blue, responsible for selected backend systems.
Business
Technological
Customer experience
In the rapidly changing market environment that defines the SmartMatch solution, having a committed team is essential. They must also be able to turn ideas into working solutions while understanding the business domain. The fireup.pro team met all of these expectations.
Aleksander Olbert
Product Owner, Trans.eu
From challenge
Key Challenges
Data from multiple sources, with varying quality
Data from multiple sources, with varying quality
The initial candidate list was generated by an external system, originally based on machine learning that analyzed chats between customers and carriers, including where they operated and what types of goods they transported. In practice, this model proved imprecise. Only when Trans.eu introduced actual freight orders as system entities, with assigned carriers and routes, did the quality of input data improve enough for filtering to become meaningful. SmartMatch had to operate on data that evolved together with the platform.
Modeling a complex decision-making process
Modeling a complex decision-making process
Choosing a carrier was not based on a single criterion, but on 21 factors, including certificates, trademarks, route matching, fleet age, previous experience with the customer, and more. Each filter could increase or decrease a carrier’s credibility. The challenge was not only to implement this logic, but also to present it in a way that was understandable for the business, so the product owner could see what was happening and why at every stage of the assessment.
Built from scratch and integrated with a growing platform
Built from scratch and integrated with a growing platform
SmartMatch was developed while Trans.eu was migrating from a monolithic architecture to microservices. The system had to integrate with both existing and newly created services, consume data from multiple sources, and deliver results to the new platform without interrupting the live production environment.
Key functional requirements
Multi-factor filtering
- selecting the best candidates from the initial list based on 21 business conditions, including carrier certificates, route compatibility, fleet age, cooperation history, trademarks, and others, with configurable weights for individual criteria.
Integration with an external recommendation service
- retrieving and normalizing data from the system that generated the initial carrier list.
Process visualization
- graphical representation of the recommendation algorithm in Camunda, available to the client’s business team.
Delivering results to the platform
- exporting a sorted recommendation list to Trans.eu’s microservices-based platform.
Access to data from multiple services
- aggregating carrier information from various sources, including certificates, activity statistics, and order history.
Through the solution
Through the solution
Architecture and approach
- Team Blue built SmartMatch from scratch as a backend service using Java and Spring. The service retrieved raw recommendations from an external system and processed each candidate through a multi-stage filtering workflow implemented and executed with Camunda.
- Camunda served two purposes. First, it executed the business workflow by evaluating carriers through consecutive filtering stages. Second, it provided a graphical visualization of the entire decision process. This allowed the team to clearly demonstrate to business stakeholders how each recommendation was produced, which filters were applied, and why a carrier received a higher or lower score—creating a transparent bridge between backend logic and product decisions.
- Carrier information, including certificates, historical data, and activity statistics, was retrieved from Trans.eu platform services. Apache Spark was used for data analysis and aggregation, while RabbitMQ handled communication between system components. Elasticsearch, Kibana, and Grafana provided logging, monitoring, and operational insights.
Delivery approach
fireup.pro worked as a dedicated backend team consisting of several Java developers and a QA engineer. The project was delivered in two-week development sprints, with full ownership of the solution—from architecture and implementation to deployment.
To the success
Technological outcomes
Built from scratch and deployed to production
SmartMatch became one of the core services within Trans.eu's new microservices platform, delivering carrier recommendations for freight orders posted by marketplace users.
Transparent decision-making process
With Camunda, every stage of the recommendation algorithm was visualized and understandable not only for engineers but also for business stakeholders. Discussions about matching logic took place using process diagrams rather than source code.
Multi-source data aggregation
The system combined carrier data from multiple Trans.eu services, including certificates, historical records, activity statistics, and customer preferences, into a unified evaluation model.
Project team
Aleksander
Backend Developer
Tech stack
Java
Spring
Camunda
Apache Spark
Amazon S3
RabbitMQ
Elasticsearch
Kibana
Grafana
Your success is our success
See how we can build a technological advantage for your company together.
We have a team that truly knows its stuff — we'll help you find a solution that works.
Conclusions & recommendations
The quality of input data determines the quality of the system
The first iterations of Trans.eu's recommendation engine, based on chat analysis, did not deliver sufficiently accurate results. Only after the platform introduced actual freight orders as system entities did SmartMatch gain a reliable foundation for filtering. Investing in high-quality source data should come before investing in advanced recommendation algorithms.
Camunda: excellent for process transparency, less suitable for high-throughput workloads
The graphical process visualization provided significant value in business communication, allowing recommendation logic to be discussed using process diagrams instead of code. However, Camunda is not designed for maximum throughput. While it performed well at Trans.eu's scale, a substantial increase in request volume would have required additional infrastructure or architectural changes.
Business process modeling as a product development tool
The "model the process before implementation" approach proved valuable beyond engineering. Camunda diagrams became living documentation, clearly showing what had been implemented, which rules represented business logic, and where edge cases occurred.
