← Boston Consulting Group Interview Insights

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

Intermediate
Jun 2026

Summary

BCG data science online assessment, notebook-style coding environment. One question on feature scaling that sounds mechanical until they ask you to justify your choices.

Questions Asked (1)

Q1

You have a DataFrame with numeric columns for age and income. Apply standard scaling to age and min-max normalization to income, then explain when you'd choose one approach over the other.

Technical Trade-offsData Modeling
Author's notes

The code part was fine, sklearn makes it pretty straightforward.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly stating the code to apply StandardScaler to age and MinMaxScaler to income, then explain the conceptual difference: standard scaling centers data around mean 0 with unit variance, while min-max normalization rescales to a fixed range [0,1]. Finally, discuss when to choose each based on algorithm assumptions, data distribution, and presence of outliers.

Pro tip: Mention that min-max normalization is sensitive to outliers because it uses the min and max, so if income has extreme values, standard scaling might be safer. Also, note that standard scaling is often preferred for algorithms that assume normally distributed data, like linear models and PCA.

1. Implement the transformations

Use sklearn's StandardScaler on the age column and MinMaxScaler on the income column. Show code or describe the steps: fit on training data, transform both train and test sets.

2. Explain standard scaling

Standard scaling subtracts the mean and divides by standard deviation, resulting in a distribution with mean 0 and variance 1. It preserves the shape of the original distribution.

3. Explain min-max normalization

Min-max normalization subtracts the minimum and divides by the range (max - min), scaling values to a fixed interval, typically [0,1]. It preserves the relationships among values but is sensitive to outliers.

4. Compare when to use each

Choose standard scaling when the algorithm assumes normally distributed data (e.g., linear regression, logistic regression, PCA) or when outliers are present but not extreme. Choose min-max when you need bounded values (e.g., neural networks with sigmoid activation) or when the algorithm is distance-based and you want to preserve zero entries (e.g., some clustering).

5. Consider data characteristics and algorithm requirements

Discuss the impact of outliers, distribution shape, and the specific algorithm's assumptions. For example, tree-based models are invariant to scaling, so it may not matter.

Key Points to Mention

  • Standard scaling (Z-score normalization) centers data at mean 0 and scales to unit variance.
  • Min-max normalization rescales to a fixed range [0,1] using min and max.
  • Standard scaling is less affected by outliers than min-max, but both can be influenced.
  • Use standard scaling for algorithms assuming Gaussian distribution (e.g., linear models, PCA).
  • Use min-max when bounded values are needed (e.g., neural networks with sigmoid/tanh).
  • Tree-based models (e.g., random forest, XGBoost) do not require scaling.

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