Boston Consulting Group Interview Questions — Page 2

Boston Consulting Group·Data Scientist·Online Assessment (OA)

May 2026

BCG data scientist live coding session on CodeSignal, basically one scenario-based question about cleaning and visualizing sales data in Python. Pretty focused, no fluff.

Given raw daily SKU-level sales data, use Python (pandas plus matplotlib or seaborn) to plot a histogram of daily revenue per SKU, detect outliers using IQR or Z-score, remove them, and then re-plot the cleaned distribution.

“The scenario itself wasn't hard to parse but I spent way too long debating IQR vs Z-score instead of just picking one and moving.”

View Post

Boston Consulting Group·Data Scientist·Online Assessment (OA)

May 2026

BCG data scientist online assessment, one of those notebook-style coding tasks where you're basically just writing production-ish ML code under a timer. Not a lot of ambiguity in the prompt, but there's still enough rope to hang yourself if you fumble the pipeline structure.

Write full Python code to train a GradientBoostingClassifier using 5-fold cross-validation, report the mean ROC-AUC score, and save the final model to disk as model.pkl.

“The ask itself is pretty clear but I second-guessed whether they wanted me to fit the final model on all the data before saving or just dump whatever came out of cross-validation.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

May 2026

BCG data scientist technical screen, basically one long scenario question about a badly overfit model with hard constraints on latency and data collection. Pretty intense for a single question but it covered a lot of ground fast.

Your model has a massive gap between training and validation error (train RMSE around 4000, validation RMSE around 9500). You can't collect more data and inference has to stay under 20ms at p95. Pick three interventions to reduce overfitting, explain why each works mechanically, and walk through a concrete experiment plan including hyperparameter grids, monitoring metrics, stopping criteria, and how you'd confirm the improvement is statistically real.

“This one took me a second to organize because there's no single right answer and the constraints actually matter.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

May 2026

BCG data scientist interview with a meaty technical case around marketing measurement. One question, but it had a lot of moving parts and I felt like I was rebuilding an analytics curriculum on the spot.

A national retail client runs weekly SKU-level marketing campaigns and has three years of historical data. How would you design an analytical approach to measure campaign lift? Walk through your choice of target variable, predictors, model(s), validation strategy, and key caveats, and explain how the findings would shape future campaigns.

“This one sprawled fast.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

May 2026

BCG data scientist technical screen, heavy on probability and Bayesian reasoning. The whole thing was basically one multi-part coin problem that escalated from posterior computation into expected value derivation.

You have a fair coin and a biased coin where P(Heads) = 0.7. You pick one uniformly at random and flip it three times. Given that the first two flips are both Heads, what is the probability you picked the biased coin?
Using the same setup, what is the probability that the third flip is also Heads, given the first two were Heads?
Still using the same coin from the earlier flips, let T be the total number of flips until you get the first Tail, counting that Tail flip. Derive E[T] given that the first two flips were both Heads.

“Classic Bayes setup.”

View Post

Boston Consulting Group·AI Engineer·Technical Phone Screen

Jun 2026

BCG AI Engineer interview with a coding question around string processing. Pretty straightforward on the surface but easy to overthink.

Given a string of letters, count how many times consecutive characters differ when treated case-insensitively.

“Took me a second to realize they just wanted adjacent comparisons, not anything fancy.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

Jun 2026

BCG data scientist interview with a code review segment over Zoom. They shared three short Python functions from a data pipeline and asked me to walk through them and suggest improvements. Not the kind of thing I'd prepped for specifically, and it showed.

Walk through these three Python functions line by line: what is each one doing and why? Then identify at least three concrete improvements across readability, efficiency, edge-case handling, or testability.

“I did okay on the 'what is it doing' part but my improvement suggestions were pretty surface level at first.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

Jun 2026

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.

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.
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?
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.
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.
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%?
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?

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

View Post

Boston Consulting Group·Data Scientist·Online Assessment (OA)

Jun 2026

BCG CodeSignal online assessment for a Data Scientist role, multiple-choice and theory block focused on ML fundamentals. Pretty standard stuff but the class-imbalance follow-up had more depth to it than I expected.

What does AUC measure, and how would you interpret a model with AUC = 0.5 versus AUC = 0.9? Also, name three techniques for handling severe class imbalance in binary classification and explain each.

“The AUC part felt fine, I knew the true/false positive tradeoff story and the 0.5 means random baseline thing.”

View Post

Boston Consulting Group·Data Scientist·Technical Phone Screen

Jun 2026

BCG data scientist interview that was basically one long technical problem about deploying a classifier on a heavily imbalanced dataset. The whole thing revolved around precision constraints and threshold selection, with a side of 'write the code too.' Pretty intense for what felt like a single question.

You have an imbalanced dataset with about 5% positive class. The product team requires test-set precision of at least 0.95 on the positive class. Walk through how you'd train a probabilistic model, find the right classification threshold on validation data, and then evaluate on the held-out test set. Report precision, recall, predicted positive count, and expected false positives if you flag 1,000 items.
If no threshold on the validation set can achieve 0.95 precision, what are two concrete strategies you'd propose, and what are the risks of each?
Write sklearn-style code to search for the optimal threshold using precision_recall_curve or a custom scorer, without leaking test data into the selection process.
How does class imbalance affect calibration, and why is optimizing ROC AUC a misleading objective when you have a hard precision constraint?

“This was the core of the whole interview.”

View Post