mediumMultiple Choice
PDE Practice Question: After deploying a model, the team notices that…
After deploying a model, the team notices that predictions are significantly different from training data distribution. What should they do?
⚠ Common exam trap
Google Cloud often tests the distinction between reactive troubleshooting (reviewing pipelines, retraining) and proactive monitoring (skew detection), tempting candidates to choose a fix like retraining instead of the monitoring solution that detects the issue first.
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
✓
Set up Vertex AI Model Monitoring for skew detection
Vertex AI Model Monitoring is specifically designed to detect skew between training data and serving data, including prediction drift. When predictions differ significantly from the training distribution, this indicates a skew or drift issue that Model Monitoring can alert on, enabling proactive investigation. Updating the endpoint or retraining without diagnosis would not address the root cause, and reviewing the pipeline alone does not provide ongoing detection.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Update the model endpoint
Why it's wrong here
Updating the model endpoint redeploys or reconfigures the serving infrastructure, which cannot reconcile a statistical divergence between live inputs and the training distribution. It is tempting because endpoint changes are the usual remedy for serving errors or stale artefacts, and would be correct if the deployed model version were outdated or misconfigured.
- ✗
Review the training data pipeline
Why it's wrong here
Reviewing the training pipeline addresses data preparation, but the stem describes live predictions diverging from the training distribution, which points to drift or a serving-versus-training skew rather than pipeline defects. It is tempting because pipeline bugs do cause distribution mismatch, and auditing it would be correct when training and serving features are built by different code paths.
- ✓
Set up Vertex AI Model Monitoring for skew detection
Why this is correct
Training-serving skew arises when live feature distributions diverge from those seen during training. Vertex AI Model Monitoring computes skew metrics by comparing production traffic against the training baseline, directly satisfying the requirement to detect distributional divergence and alert the team before predictions degrade further.
- ✗
Retrain the model with new data
Why it's wrong here
Retraining with new data changes the model's learned parameters, but the stem asks what to do about observed distribution divergence, which requires detecting and diagnosing drift before any retraining decision. It is tempting because retraining is the eventual remedy once drift is confirmed, and would be correct after monitoring establishes that the input distribution has genuinely shifted.
Go deeper
Related to this question
About these practice questions
Courseiva writes every PDE question from scratch — 747 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. 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.