Transforming old software into new solutions is a complex task, especially when test coverage needs to be optimized. One team was inspired by soccer analyses and uses process mining to uncover the most important workflows and throughput variants from real production data.
In this episode, I talk to Sven Braxein and Athanasios Kallinikidis about how they use soccer analytics to improve test coverage in a software project. The idea arose spontaneously when Thanos was editing clips for his soccer team in the office. This turned into an exciting dialog: Can we learn from real production data the way we learn from sports statistics? We talk about process mining, numerators and denominators when it comes to test coverage, surprising findings and the difference between opinion and knowledge.
"If it's already possible for someone to say 'I'm learning from my game for training' in a soccer league, then it must also be possible for us in IT to learn something from our production data for testing." - Sven Braxein
Athanasios is 28 years old, has Greek roots and was born and raised in the Esslingen district near Stuttgart. He graduated from the Technical University of Munich with a Bachelor's and Master's degree in Management & Technology, specializing in Computer Science. An internship at Celonis sparked his passion for process mining and laid the foundation for his career focus. Today, he drives the further development of the process mining platform at Mercedes-Benz Leasing Germany as Product Owner in the Platform Team Business-IT. Athanasios is a passionate soccer fan and has discovered coaching as a second vocation. As a working student, he worked in match analysis for FC Bayern Munich II in the 3rd division. He is currently transferring the analysis skills he learned there to his work with his local men's team.
Sven Braxein is the founder and managing director of TestGilde GmbH, which was founded in 2008.He has been working for 30 years in IT project consulting with a focus on software quality assurance and testing. He specializes in test management for large-scale projects and in the design and implementation of company-wide test management structures.If the topic of quality and testing did not already exist, it would have to be invented for him.
Software testing often sounds like a dry topic. But the Software Testing podcast features people who are passionate about it. In this episode, Richie shows how much excitement there is in the world of quality and test processes. His guests are Athanasios Kallinikidis, soccer coach and software expert, and Sven Braxein, test professional from a large IT project. Their story begins with a completely different discipline: soccer. What can quality assurance learn from soccer analysis? Quite a lot.
Sven and Athanasios' project has one major goal: to replace the old software with a new one. That sounds simple, if it weren't for the question of how to test and what needs to be tested at all. At the beginning, the reproach comes: "Your coverage in regression testing is too poor." But what should actually be covered? How do you recognize which tests really count? Sven Braxein asks the classic question of percentage calculation: What is the denominator and what is the numerator? The answer in everyday life is often missing.
Suddenly, by chance, an idea emerges. Athanasios Kallinikidis is sitting at his laptop in the office editing video clips for his soccer team. He uses an inexpensive camera and special software that tracks game situations and compiles statistics. Every player, every run, every shot is recorded. Sven Braxein listens and thinks: "If it even works in amateur soccer, why not in software projects? In soccer, you learn from real match data for training. In an IT project, it should also be possible to learn something for testing from the real usage data of the software.
The conversation leads back to professional life. Athanasios works with process mining tools in the company. Process mining combines classic process analysis from consulting with the possibilities of data mining. Every click, every action in the system leaves traces. With process mining, you can summarize these traces and see exactly which processes are actually taking place.
There are 160 different workflows in the software project. Each workflow represents a business process, such as creating, terminating or changing a contract. Using the process data from the production environment, the tool shows which workflows and which runs occur most frequently in the real world. A workflow is run through up to 600 different times. The statistics can evaluate which ten runs together already cover 80 percent of real usage.
This analysis enables a new test model: instead of testing according to feelings or old assumptions, the test cases are selected according to actual real-world usage data. The top 10 workflows are thoroughly tested. The rarely used ones can be neglected.
The soccer comparison helps. "More knowledge, less opinion," says Athanasios Kallinikidis. In training, there is no longer a discussion about who has run the most. The numbers show it. It works the same way in software projects. Test coverage is based on real needs, no longer on assumptions.
The podcast makes it clear that the results are sometimes surprising. Workflows that tend to be neglected on the test list are the most important in production. Conversely, the analysis also shows which processes are perhaps no longer needed and where test resources should be better deployed.
However, this method also has its limits. At some point, the effort involved becomes too high if you want to include all small systems and databases. The benefit decreases. Routine is crucial: regular evaluation and an automated process ensure that the method helps in the long term.
The goal remains: Test coverage should be designed in such a way that it not only delivers numbers, but also tests the right thing. Just like in soccer training: you learn the most from real game situations, not from assumptions from the sidelines.
The link between soccer and software testing shows how fresh ideas from other areas of life can help to improve IT processes. If you use real data, you quickly find out what really needs to be tested. This way, testing is not a guessing game, but a real quality improvement.
And sometimes, just like in soccer, it's the little things that make the difference between winning and losing - it's best if you've seen them beforehand.