Google·Software Engineer·Technical Phone Screen·Intermediate

IntermediatePrefer not to say

May 2026

Summary

Google data or ML role, one technical screen that came down to a pretty foundational preprocessing question. Short and to the point, nothing fancy.

Questions Asked (1)

What is the difference between normalization and standardization, and when would you choose one over the other?

Technical Trade-offsAlgorithms & Data Structures

Author's notes

Seems basic until you actually have to explain the tradeoffs out loud.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly defining both terms and their mathematical transformations, then contrast their effects on data distribution and scale. Discuss practical scenarios where each is preferred, emphasizing the impact on model performance and interpretability. Conclude with a decision framework based on data characteristics and algorithm requirements.

Pro tip: Mention that standardization is generally preferred for algorithms assuming Gaussian distributions (e.g., linear models, PCA), while normalization is better for distance-based algorithms (e.g., KNN, neural networks) and when data has bounded ranges. Also note that the choice can affect convergence speed and model accuracy, so it's often worth experimenting with both.

1. Define Normalization

Explain that normalization (min-max scaling) rescales features to a fixed range, typically [0, 1], by subtracting the minimum and dividing by the range. It preserves the shape of the original distribution but is sensitive to outliers.

2. Define Standardization

Explain that standardization (Z-score normalization) transforms features to have zero mean and unit variance by subtracting the mean and dividing by the standard deviation. It centers the data and scales it, making it less sensitive to outliers but assuming a Gaussian-like distribution.

3. Compare Effects on Data

Highlight that normalization bounds values, which is useful when features have different units or when algorithms require inputs in a specific range. Standardization preserves the shape but shifts and scales, which is beneficial for algorithms that assume normally distributed data.

4. When to Choose Normalization

Choose normalization when features have varying scales and the algorithm is sensitive to the magnitude of values (e.g., KNN, K-means, neural networks) or when you need bounded inputs (e.g., image processing). Also use when data does not follow a Gaussian distribution.

5. When to Choose Standardization

Choose standardization when the algorithm assumes Gaussian distribution (e.g., linear regression, logistic regression, PCA) or when features have outliers that should not dominate the scaling. It is also preferred when the data has varying scales but you want to preserve the relative differences.

Key Points to Mention

Normalization scales to a fixed range [0,1] (or [-1,1]), while standardization scales to mean=0, std=1.
Normalization is sensitive to outliers; standardization is more robust to outliers.
Standardization is preferred for algorithms that assume Gaussian distribution (e.g., linear models, PCA).
Normalization is preferred for distance-based algorithms (e.g., KNN, K-means) and neural networks.
The choice can impact model convergence and performance; sometimes both are tested.
Consider the nature of the data: if bounded and not Gaussian, normalize; if Gaussian-like, standardize.

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