← Boston Consulting Group Interview Insights

Boston Consulting Group·Data Scientist·Technical Phone Screen·Senior

SeniorPrefer not to say
Jun 2026

Summary

BCG data scientist interview with a deep end-to-end ML case on credit card default modeling. The whole thing was basically one long problem broken into six parts, and it got pretty technical pretty fast. Walked away feeling okay about most of it but the calibration math under sampling shift was genuinely stressful in the moment.

Questions Asked (6)

Q1

How would you frame a 90-day credit card default prediction problem at the account-month level, including how you define the label, what features are available at scoring time, and how you set up a temporal train/validation split to avoid leakage? Name three leakage risks specific to credit cards.

Data ModelingTechnical Trade-offs
Author's notes

I started with the label definition and felt okay there (observation window, prediction horizon, what date counts as 'default').

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining the prediction problem clearly: predict probability of default within 90 days at the account-month level, using only data available at the scoring time. Then outline the label definition, feature engineering with temporal considerations, and a robust temporal validation strategy to prevent leakage. Finally, highlight three credit-card-specific leakage risks and how to mitigate them.

Pro tip: Emphasize that the label window and feature window must be strictly separated, and that any aggregation (e.g., average spend) must be computed only on data prior to the scoring date. This shows practical awareness of temporal leakage.

1. Define the prediction problem and label

Specify the unit of analysis (account-month), the prediction horizon (90 days), and the label as whether the account defaults within 90 days after the scoring date. Clarify the definition of default (e.g., 90+ days past due) and ensure it aligns with business objectives.

2. Identify features available at scoring time

List features that are known at the scoring date, such as account attributes, transaction history up to that point, and payment behavior. Exclude any post-scoring information and ensure all features are computed using only data prior to the scoring date.

3. Set up temporal train/validation split

Use a time-based split (e.g., train on earlier months, validate on later months) to mimic real-world deployment. Ensure that the validation period is strictly after the training period and that no future data leaks into training.

4. Address leakage risks specific to credit cards

Identify and mitigate three common leakage risks: (1) using future payment behavior (e.g., payments made after scoring date) as features, (2) including account status changes that occur after scoring, and (3) aggregating transaction data over a window that extends beyond the scoring date.

5. Validate and monitor

Evaluate model performance on the temporal validation set using appropriate metrics (e.g., AUC, KS). Plan for ongoing monitoring to detect drift and ensure the model remains leakage-free in production.

Key Points to Mention

  • Label definition: default within 90 days after scoring date, aligned with business definition (e.g., 90+ days past due).
  • Feature availability: only use data known at scoring time; avoid any post-scoring information.
  • Temporal split: train on past data, validate on future data to simulate real deployment.
  • Leakage risk 1: using future payment behavior (e.g., payments made after scoring) as features.
  • Leakage risk 2: including account status changes (e.g., closure, delinquency) that occur after scoring.
  • Leakage risk 3: aggregating transaction data over a window that extends beyond the scoring date.

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

Q2

With only 2% default prevalence in production, which evaluation metrics would you choose and why? How do you pick an operating point that actually aligns with business goals?

Product Analytics & MetricsTechnical Trade-offs
Author's notes

PR-AUC was my first answer and I think that landed fine.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging the class imbalance and its impact on metric choice, then recommend precision-recall based metrics like PR-AUC and F1 over ROC-AUC. Explain how to select an operating point by translating business costs and benefits into a threshold that maximizes expected value, not just statistical performance.

Pro tip: Always tie the operating point to a concrete business KPI (e.g., cost per default avoided or expected profit) and mention that you would validate the threshold on a holdout set with temporal stability checks.

1. Assess the imbalance and metric limitations

Explain why accuracy and ROC-AUC can be misleading with 2% prevalence, and why precision-recall metrics are more informative.

2. Choose appropriate evaluation metrics

Recommend PR-AUC, F1, precision@k, recall@k, and lift, and justify each based on the business context (e.g., cost of false positives vs. false negatives).

3. Map business objectives to costs and benefits

Quantify the cost of a false positive (e.g., unnecessary intervention) and false negative (e.g., default loss), and express the net benefit as a function of the threshold.

4. Select the operating point

Choose the threshold that maximizes expected business value (e.g., profit, cost savings) or meets a required recall/precision constraint, and validate it on out-of-time data.

5. Monitor and adjust

Set up monitoring for data drift and business metric changes, and plan to re-evaluate the threshold periodically.

Key Points to Mention

  • Class imbalance makes accuracy and ROC-AUC misleading; precision-recall curve and PR-AUC are more sensitive to the minority class.
  • F1 score balances precision and recall but assumes equal cost; consider cost-sensitive metrics like expected cost or profit.
  • Precision@k and recall@k are useful when the business can only act on top k predictions (e.g., limited resources).
  • The operating point should be chosen by optimizing a business objective function, not by maximizing a statistical metric alone.
  • Threshold selection must consider the trade-off between false positives and false negatives, guided by their relative costs.
  • Validate the chosen threshold on a holdout set and monitor for drift to ensure it remains aligned with business goals over time.

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

Q3

You downsample negatives so your training set has a 1:3 positive-to-negative ratio, giving a training prevalence of 25% versus the true 2%. If the model outputs a score of 0.60 for an account, derive the calibrated population probability using prior-probability correction and compute the numeric result.

Data ModelingTechnical Trade-offsAlgorithms & Data Structures
Author's notes

This was the part I was most nervous about and honestly it showed.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, clarify that the model's output is a probability under the downsampled distribution, so it must be adjusted to reflect the true population prevalence. Then, apply the prior-probability correction formula: p_true = (p_model * π_true) / (p_model * π_true + (1 - p_model) * (1 - π_true)), where π_true is the true prevalence (0.02). Finally, plug in p_model = 0.60 and compute the numeric result.

Pro tip: Always state the assumption that the model is well-calibrated on the training distribution and that the likelihood ratios are preserved. This shows you understand the underlying assumptions and can communicate them clearly to stakeholders.

1. Identify the given prevalences

Note the training prevalence (π_train = 0.25) and the true population prevalence (π_true = 0.02). Recognize that the model's output is a probability under the training distribution.

2. State the calibration formula

Use the prior correction formula: p_true = (p_model * π_true) / (p_model * π_true + (1 - p_model) * (1 - π_true)). Explain that this adjusts for the different base rates.

3. Plug in the values

Substitute p_model = 0.60, π_true = 0.02, and (1 - π_true) = 0.98 into the formula.

4. Compute the result

Calculate the numerator and denominator: numerator = 0.60 * 0.02 = 0.012; denominator = 0.012 + 0.40 * 0.98 = 0.012 + 0.392 = 0.404. Then p_true = 0.012 / 0.404 ≈ 0.0297.

5. Interpret the result

The calibrated probability is approximately 2.97%, which is much lower than the model's raw score of 60%, reflecting the true rarity of positives in the population.

Key Points to Mention

  • The model's output is a probability under the downsampled distribution, not the true population.
  • The prior correction formula adjusts for different base rates and assumes the model's likelihood ratios are preserved.
  • The training prevalence is 25% (1:3 ratio), while the true prevalence is 2%.
  • The calculation yields a calibrated probability of about 2.97%.
  • This correction is crucial when deploying models trained on balanced or downsampled data to real-world imbalanced settings.
  • Communicate the result clearly, emphasizing the significant difference from the raw score.

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

Q4

Outreach costs $2 per account, and a successful intervention on a defaulting account has a 30% chance of averting a $150 loss. Derive the breakeven probability threshold for contacting an account, and apply it to the calibrated probability from the previous part.

Product Analytics & MetricsPricing & Monetization
Author's notes

Expected value setup: contact if p x 0.30 x 150 > 2, so p* = 2/45 = roughly 0.044.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

First, set up the expected value equation for contacting an account: cost = $2, benefit = p * $150, where p is the probability that the intervention averts the loss. Solve for the breakeven p where expected benefit equals cost. Then compare this threshold to the calibrated probability from the previous part to decide whether to contact the account.

Pro tip: Always clarify that the 30% success rate is the intervention's effectiveness conditional on contacting a defaulting account, and that the calibrated probability is the likelihood the account will default. The overall probability of averting loss is the product of these two, so the breakeven condition is on the calibrated default probability, not on the intervention success rate alone.

1. Define variables and assumptions

Let p be the calibrated probability that the account defaults. The intervention, if applied, has a 30% chance of averting the $150 loss. Assume the intervention only matters if the account defaults; otherwise, no loss occurs and no benefit is gained.

2. Set up expected value equation

Expected benefit of contacting = p * 0.30 * $150. Expected cost = $2. Breakeven when expected benefit equals cost: p * 0.30 * $150 = $2.

3. Solve for breakeven probability

Compute p = $2 / (0.30 * $150) = $2 / $45 ≈ 0.0444, or about 4.44%. This is the minimum calibrated default probability required to justify the outreach cost.

4. Apply to calibrated probability

Compare the calibrated probability from the previous part to 4.44%. If the calibrated probability is greater than 4.44%, contacting the account has positive expected value; if less, it does not.

5. Interpret and communicate

State the decision clearly: contact accounts with calibrated default probability above 4.44%. Discuss sensitivity: if the intervention success rate or loss amount changes, the threshold shifts.

Key Points to Mention

  • Expected value calculation: cost vs. probabilistic benefit
  • Breakeven probability formula: p = cost / (success_rate * loss_amount)
  • Numerical result: p ≈ 4.44%
  • Comparison with calibrated probability to make a contact decision
  • Assumption that intervention only affects defaulting accounts
  • Sensitivity analysis: how threshold changes with different parameters

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

Q5

Design a time-based cross-validation scheme and a backtest that checks model stability across different regions and macro environments. How would you monitor calibration and concept drift after deployment, and what would you do if the true default rate dropped from 2% to 1%?

A/B Testing & ExperimentationSystem DesignRoot Cause Analysis
Author's notes

Talked through walk-forward validation with expanding windows and held out the most recent months as a backtest.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Structure your answer around three phases: design, validation, and monitoring. Start by explaining how you would build a time-based cross-validation scheme that respects temporal order and incorporates regional and macro-economic stratification. Then describe a backtest that evaluates model stability across these dimensions, and finally outline a post-deployment monitoring plan for calibration and drift, including a specific response to a drop in default rate.

Pro tip: Emphasize that a drop in default rate is not necessarily a model failure—it could be a true improvement in credit quality. Propose a root-cause analysis to distinguish between data drift, concept drift, and genuine population change before taking action.

1. Design Time-Based Cross-Validation

Use expanding or rolling window cross-validation that respects temporal order, ensuring training data always precedes validation data. Incorporate stratification by region and macro-economic regime (e.g., recession vs. expansion) to test model performance across diverse conditions.

2. Backtest for Stability

Backtest the model on historical data segmented by region and macro environment, comparing performance metrics (e.g., AUC, calibration) across segments. Use statistical tests to assess if performance differences are significant and identify regions or regimes where the model underperforms.

3. Monitor Calibration and Drift Post-Deployment

Implement ongoing monitoring of calibration (e.g., calibration plots, Brier score) and data/concept drift (e.g., population stability index, Kolmogorov-Smirnov tests). Set up alerts for significant deviations and regularly retrain or recalibrate as needed.

4. Respond to Default Rate Drop

If the true default rate drops from 2% to 1%, first verify the drop is real and not due to data issues. Then analyze whether the model's predicted probabilities remain calibrated; if not, recalibrate using recent data. Investigate if the drop is due to a change in population or economic conditions, and adjust the model or strategy accordingly.

Key Points to Mention

  • Temporal validation to prevent look-ahead bias
  • Stratification by region and macro-economic indicators (e.g., GDP growth, unemployment)
  • Calibration metrics: reliability diagrams, Brier score, expected calibration error
  • Drift detection: PSI, KS test, and monitoring of feature distributions
  • Root-cause analysis for default rate change: data quality, population shift, concept drift
  • Business impact: adjusting thresholds or strategies in response to calibration changes

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

Q6

What fairness checks would you run on this model, and what operational guardrail would you add to prevent the model from over-throttling credit limits due to uncertainty? How can your sampling choices distort fairness metrics if left uncorrected?

Technical Trade-offsCross-functional AlignmentProduct Analytics & Metrics
Author's notes

Named equal opportunity across demographic groups and calibration parity as my two checks.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Structure your answer around three pillars: fairness checks (e.g., demographic parity, equal opportunity, calibration across groups), an operational guardrail (e.g., human-in-the-loop review or uncertainty-based abstention for credit limit changes), and sampling bias correction (e.g., reweighting or stratified sampling). Emphasize the trade-off between fairness and business metrics, and how you would communicate this to cross-functional stakeholders.

Pro tip: Frame fairness not as a one-time check but as a continuous monitoring system with automated alerts, and propose a 'fairness budget' that quantifies acceptable disparity, making trade-offs explicit to product and compliance teams.

1. Define fairness metrics and groups

Identify protected groups (e.g., by race, gender, age) and select appropriate fairness metrics such as demographic parity, equalized odds, or predictive parity, aligning with regulatory and business context.

2. Run fairness audits and diagnose disparities

Compute metrics across groups, test for statistical significance, and investigate root causes (e.g., label bias, feature proxies) using tools like Fairlearn or AIF360.

3. Design operational guardrails

Implement uncertainty-aware thresholds (e.g., only auto-adjust limits when model confidence exceeds a threshold), human review for borderline cases, and a kill switch to revert to a safe default policy.

4. Correct sampling distortions

Apply reweighting, stratified sampling, or post-stratification to ensure fairness metrics reflect the true population, and validate with holdout sets that mimic deployment distribution.

5. Monitor and iterate

Set up continuous fairness monitoring with alerts for drift, and establish a feedback loop with cross-functional teams to update the model and guardrails as needed.

Key Points to Mention

  • Demographic parity, equal opportunity, and calibration as key fairness metrics
  • Uncertainty quantification (e.g., Bayesian methods, conformal prediction) to avoid over-throttling
  • Human-in-the-loop for high-stakes decisions like credit limit changes
  • Sampling bias: oversampling/undersampling can distort group representation and fairness estimates
  • Reweighting or stratified sampling to correct for sampling bias
  • Trade-offs between fairness and business objectives, and how to communicate them

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