mediumMultiple Choice
PDE Practice Question: A company has a production model deployed on…
A company has a production model deployed on Vertex AI that shows declining accuracy over time. The model uses features from a BigQuery feature store. The data science team suspects data drift. What is the most efficient way to monitor and detect drift for this model?
⚠ Common exam trap
Google Cloud often tests the distinction between monitoring for data drift (which requires distribution comparison) and monitoring for operational metrics (like latency or error rates), leading candidates to confuse Cloud Monitoring dashboards with drift detection.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Enable Vertex AI Model Monitoring on the endpoint to automatically detect skew and drift
Vertex AI Model Monitoring is purpose-built for detecting feature skew and drift in production models. It automatically compares the distribution of prediction request data against training data statistics, alerting when significant divergence occurs — this is the most efficient and integrated approach for a Vertex AI endpoint.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Enable Vertex AI Model Monitoring on the endpoint to automatically detect skew and drift
Why this is correct
Vertex AI Model Monitoring attaches to the deployed endpoint and continuously compares incoming prediction feature distributions against the training baseline, automatically flagging training-serving skew and drift. This satisfies the drift-detection requirement without building custom BigQuery comparison pipelines.
- ✗
Periodically export training data and production data to CSV and compare distributions manually
Why it's wrong here
Manual CSV exports and distribution comparisons are labour-intensive and lag behind production traffic, so drift is detected late. Vertex AI Model Monitoring computes drift metrics automatically against BigQuery feature data, which is the efficient choice here.
- ✗
Create a scheduled retraining pipeline that runs weekly
Why it's wrong here
Retraining weekly addresses drift symptoms but never detects or reports drift itself, so the team cannot confirm the cause or tune thresholds. Vertex AI Model Monitoring measures feature and prediction drift directly, which is what the question asks for.
- ✗
Set up Cloud Monitoring dashboards to track prediction request volumes and error rates
Why it's wrong here
Request volume and error rates are operational metrics; they reveal traffic and failures, not changes in feature or prediction distributions. Vertex AI Model Monitoring computes drift against a baseline, which is the mechanism needed to detect data drift.
Go deeper
Related to this question
About these practice questions
One of 747 original PDE practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This PDE practice question is part of Courseiva's free Google Cloud certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the PDE exam.