Ml Functional Performance Metrics
ML functional performance metrics are quantitative measures used to evaluate how effectively a machine learning model performs its intended task.
“A set of measures that relate to the functional correctness of an ML model.”
(Definition reproduced from the ISTQB Glossary. Copyright belongs to ISTQB.)
View the complete ISTQB GlossaryWhat is Ml Functional Performance Metrics?
Examples include accuracy, precision, recall, F1-score, ROC-AUC, and other statistical measures depending on the ML problem.
Model evaluation: Metrics quantify prediction quality.
Problem-specific: Different ML tasks require different metrics.
Decision support: Metrics guide model selection and improvement.
Real World Example
A team deploys a fraud-detection ML model and reports that it responds to each transaction check in 50 milliseconds, comfortably within their performance budget, then declares the model production-ready.
Fast response time says nothing about whether the model's predictions are actually correct — the same model could answer in 50ms while being wrong about which transactions are fraudulent most of the time, and "performance" for an ML model can confusingly mean either speed or prediction correctness.
The tester evaluates Ml functional performance metrics specifically: measuring precision (of transactions flagged fraudulent, how many really are), recall (of all actual fraud, how much was caught), and F1 score on a held-out labeled dataset, entirely separate from any latency or resource measurement.
The evaluation reveals the model catches only 60% of actual fraud cases despite its fast response time — a defect completely invisible to a speed-focused dashboard, caught only because Ml functional performance metrics were measured as their own distinct concern from response time.
Practice Questions
Question 1
A fraud-detection model responds to each check in 50ms, well within budget, and the team declares it production-ready. Later, the model is found to only catch 60% of actual fraud cases. What was missing from the original evaluation?
Question 2
Which metric would be classified under ML functional performance metrics for a classification model?
Test your knowledge with real ISTQB-style questions
You’ve learned Ml Functional Performance Metrics. Test your understanding with topic-specific questions in our Mock Exams.