mediumMultiple Select
MLA-C01 Practice Question: A data science team detects that a deployed…
A data science team detects that a deployed model's prediction accuracy is degrading over time due to concept drift. They need to implement a retraining strategy. Which THREE actions are recommended best practices for handling concept drift?
⚠ Common exam trap
Test-takers frequently confuse 'detecting drift' with 'responding to drift' and incorrectly choose automatic rollback (Option A) as a best practice, when in reality rollback is a risky operation that should be evaluated carefully, not automated blindly.
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
✓
Monitor prediction quality using ground truth labels when available.
Option B is correct because continuously monitoring prediction quality against ground truth labels (e.g., via SageMaker Model Monitor's ModelQuality baseline and CloudWatch metrics) is the only way to quantify real accuracy degradation and confirm concept drift rather than mere data drift. Option D is correct because SageMaker Pipelines can orchestrate incremental retraining workflows that incorporate newly labeled data, allowing the model to adapt to the changed concept without rebuilding the entire pipeline manually. Option E is correct because SageMaker Model Monitor detects drift (data drift, model quality drift) and can emit CloudWatch alarms that trigger a retraining pipeline, closing the detect-to-retrain loop automatically. Option A is not recommended as a primary strategy because rolling back to an old model does not address the underlying concept drift and the previous version will also degrade on the new distribution. Option C is not recommended because fixed-schedule retraining ignores actual performance signals, wasting compute when drift is absent and lagging when drift is rapid.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Automatically roll back to a previous model version upon drift detection.
Why it's wrong here
Rolling back restores an older model trained on stale data, so it cannot capture the new concept relationship and accuracy stays degraded. It is tempting because rollback is the right response to a bad deployment that introduced a regression, not to gradual concept drift in the underlying data.
- ✓
Monitor prediction quality using ground truth labels when available.
Why this is correct
Monitoring prediction quality against ground truth labels detects actual accuracy degradation, not just input drift. This satisfies the stem's concept drift scenario by measuring whether the model's outputs still match reality as the relationship between features and target changes.
- ✗
Retrain the model on a fixed schedule regardless of performance.
Why it's wrong here
A fixed schedule retrains on stale or irrelevant data and wastes compute when no drift has occurred, ignoring the detected performance signal. It is tempting because scheduled retraining is valid when drift is known to be seasonal or predictable, where a calendar cadence matches the data cycle.
- ✓
Incrementally update the model with new data using SageMaker Pipelines.
Why this is correct
SageMaker Pipelines orchestrates incremental retraining, feeding fresh data to the model so it adapts to the shifted input-output relationship that defines concept drift. This directly satisfies the stem's requirement for a retraining strategy, automating repeatable updates rather than relying on static, one-off training runs.
- ✓
Use SageMaker Model Monitor to detect drift and trigger retraining.
Why this is correct
SageMaker Model Monitor continuously evaluates live endpoint data against a baseline, emitting CloudWatch metrics when drift exceeds thresholds. Those alarms can invoke a Lambda function or Step Functions workflow that triggers automatic retraining, directly satisfying the stem's requirement to detect concept drift and act on it without manual intervention.
Go deeper
Related to this question
About these practice questions
One of 665 original MLA-C01 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 MLA-C01 practice question is part of Courseiva's free Amazon Web Services 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 MLA-C01 exam.