← Google Interview Insights

Google·Data Scientist·Technical Phone Screen·Intermediate

IntermediatePrefer not to say
May 2026

Summary

Interviewed at Google, got hit with a stats/ML eval question that seems basic but I fumbled the explanation more than I expected.

Questions Asked (1)

Q1

What does it mean when a model's AUC is 0.5, and why does that matter?

Product Analytics & MetricsTechnical Trade-offs
Author's notes

I knew the textbook answer, a model at 0.5 is basically guessing randomly, no better than a coin flip.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining AUC as the probability that the model ranks a random positive instance above a random negative instance. Then explain that AUC=0.5 means the model's rankings are no better than random chance, and discuss why this matters for model evaluation, business impact, and next steps.

Pro tip: Emphasize that AUC=0.5 doesn't necessarily mean the model is useless—it could indicate a data issue, a need for better features, or that the problem is inherently unpredictable. Always investigate before discarding the model.

1. Define AUC

Explain that AUC (Area Under the ROC Curve) measures the model's ability to discriminate between positive and negative classes across all thresholds. AUC=0.5 means the model has no discriminative power.

2. Interpret AUC=0.5

State that AUC=0.5 indicates the model performs no better than random guessing. The ROC curve would be a diagonal line from (0,0) to (1,1).

3. Explain why it matters

Discuss the implications: the model is not useful for ranking or classification, may lead to poor business decisions, and signals a need for improvement or investigation.

4. Consider possible causes

Mention potential reasons: poor feature engineering, data leakage, incorrect labeling, or the problem being inherently random. Also note that AUC=0.5 can occur if the model is overfitting or underfitting.

5. Recommend next steps

Suggest actions: re-evaluate data quality, try different models, add features, or consider alternative metrics if AUC is not appropriate for the problem.

Key Points to Mention

  • AUC=0.5 means the model's predictions are equivalent to random chance.
  • It indicates no discriminative ability between positive and negative classes.
  • This can be due to data issues, model issues, or an inherently unpredictable target.
  • It matters because the model provides no value for decision-making or ranking.
  • Investigate before concluding: check for data leakage, label errors, or feature problems.
  • Consider alternative metrics if the problem is imbalanced or if ranking is not the goal.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.