Risk-Based Testing and Prioritization

Meta (formerly Facebook), the social media and technology conglomerate, provides one of the most thoroughly documented implementations of ML-based test selection at scale. According to a Meta engineering publication from 2020, the company deployed a predictive test selection system across its trunk-based development model. The system uses a gradient-boosted decision-tree model trained on features derived from previous code changes and test outcomes to estimate the probability of each test detecting a regression. After more than a year of production deployment, the system catches more than 99.9% of all regressions while running only one-third of transitively dependent tests, doubling infrastructure efficiency and requiring little to no manual tuning as the codebase evolves.

In the telecommunications sector, Nokia undertook a multi-year research initiative to introduce machine learning software defect prediction into the system-level testing of its 5G platform. A 2025 peer-reviewed study published in the Journal of Systems and Software by Madeyski and Stradowski validated a lightweight alternative to traditional defect prediction, demonstrating feasibility using existing Nokia 5G test process data. The research, presented at the IEEE/ACM International Conference on Software Engineering in 2023, found that return-on-investment and benefit-cost ratio calculations confirmed the approach can deliver positive monetary impact in both lightweight and advanced deployment scenarios. Nokia practitioners noted that the solution required no additional cost beyond engineering hours to implement, reinforcing the low barrier to entry for organizations with mature data collection practices.