← Experian Interview Insights

Experian·Data Scientist·Technical Phone Screen·Intermediate

Intermediate
May 2026

Summary

Technical screen for a data scientist role at Experian DataLabs. The interviewer went deep on a binary classification project I had on my resume, mixing conceptual ML questions with stuff that felt almost like multiple choice exam material. Pretty rigorous for a phone screen.

Questions Asked (5)

Q1

How does PCA reduce dimensionality, and should you apply L2 normalization before training? Also, what's the difference between standardizing each column versus normalizing each row, and when does that distinction actually matter?

Technical Trade-offsAlgorithms & Data Structures
Author's notes

I got the eigenvector/explained variance part fine but then fumbled the normalization distinction harder than I expected.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by explaining PCA's core mechanism: it finds orthogonal linear combinations of features that maximize variance, effectively projecting data onto a lower-dimensional subspace. Then address the normalization question by distinguishing between L2 normalization (often used in text/embedding contexts) and standardization (critical for PCA when features have different scales). Finally, clarify the difference between column-wise standardization and row-wise normalization, emphasizing when each is appropriate based on the data type and modeling goal.

Pro tip: Mention that PCA is sensitive to feature scales, so standardization is almost always necessary unless all features are already on the same scale. Also, note that L2 normalization before PCA can be useful for sparse data like text, but it changes the geometry and may not be ideal for all cases.

1. Explain PCA's dimensionality reduction

Describe how PCA computes the covariance matrix, finds eigenvectors (principal components), and projects data onto the top k components that capture the most variance. Emphasize that this reduces dimensionality while preserving as much information as possible.

2. Address L2 normalization before training

Clarify that L2 normalization scales each sample to unit norm, which is common in text classification or embedding spaces. Discuss whether it's appropriate: it can help with cosine similarity but may remove magnitude information that PCA could leverage.

3. Differentiate column-wise standardization vs row-wise normalization

Explain that standardizing each column (feature) subtracts mean and divides by standard deviation, making features comparable. Normalizing each row (sample) scales the sample vector to unit norm, often used in text or when sample magnitude is irrelevant.

4. Discuss when the distinction matters

Highlight scenarios: column-wise standardization is crucial for PCA when features have different units/scales. Row-wise normalization matters when the analysis should be invariant to sample magnitude, e.g., in document clustering or when using cosine similarity.

5. Summarize with practical recommendations

Conclude that for PCA, standardizing columns is typically recommended. L2 normalization before PCA is context-dependent and should be applied cautiously, often after standardization if at all.

Key Points to Mention

  • PCA maximizes variance and requires features to be on the same scale for meaningful results.
  • Standardization (column-wise) is essential when features have different units or variances.
  • L2 normalization (row-wise) scales samples to unit norm, useful for text data but can distort PCA if applied without thought.
  • The distinction matters in domains like NLP where row normalization is common, versus tabular data where column standardization is key.
  • Applying L2 normalization before PCA can be beneficial for sparse data but may not be necessary for dense data.
  • Always consider the downstream task: if using cosine similarity, row normalization might be preferred; if using Euclidean distance, column standardization is often better.

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

Q2

Derive the gradient for logistic regression using backpropagation, then explain how backprop generalizes to a multi-layer neural network.

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

Honestly the derivation part went okay.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by deriving the gradient for logistic regression using the chain rule, explicitly showing the forward pass (linear combination, sigmoid activation) and backward pass (error term, weight gradients). Then generalize to multi-layer networks by introducing hidden layers, activation functions, and the recursive application of the chain rule through backpropagation, emphasizing the role of the Jacobian and error propagation.

Pro tip: Connect the derivation to practical considerations like numerical stability (e.g., using log-sum-exp trick) and mention how frameworks like TensorFlow/PyTorch automate this, showing awareness of real-world implementation.

1. Define logistic regression model

State the model: z = w^T x + b, output y_hat = sigmoid(z). Define the loss function (binary cross-entropy) for a single example: L = -[y log(y_hat) + (1-y) log(1-y_hat)].

2. Compute gradients via chain rule

Derive dL/dz = y_hat - y (a key simplification). Then compute dL/dw = (y_hat - y) x and dL/db = y_hat - y. Show the steps clearly.

3. Extend to multi-layer networks

Introduce a network with L layers, each with weights W^l, biases b^l, and activation functions a^l. Define forward propagation: z^l = W^l a^{l-1} + b^l, a^l = f^l(z^l).

4. Derive backpropagation recursively

Define error term δ^l = ∂L/∂z^l. For output layer, δ^L = ∇_a L ⊙ f'(z^L). For hidden layers, δ^l = (W^{l+1}^T δ^{l+1}) ⊙ f'(z^l). Then gradients: ∂L/∂W^l = δ^l (a^{l-1})^T, ∂L/∂b^l = δ^l.

5. Summarize and discuss generalization

Explain that backprop is a general algorithm for computing gradients in any computational graph, and logistic regression is a special case with no hidden layers. Mention computational efficiency and modularity.

Key Points to Mention

  • Chain rule and its recursive application
  • The simplification dL/dz = y_hat - y for sigmoid + cross-entropy
  • Role of activation function derivatives (e.g., sigmoid, ReLU) in backprop
  • Matrix/vector notation for efficient batch computation
  • Vanishing/exploding gradients and mitigation strategies (e.g., ReLU, batch norm)
  • Connection to automatic differentiation in modern frameworks

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

Q3

What baseline models did you compare your logistic regression against, and what drove the decision to go with logistic regression over the alternatives?

Technical Trade-offsProduct Analytics & Metrics
Author's notes

This was the easiest question of the bunch.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly stating the baseline models you compared (e.g., dummy classifier, regularized logistic regression, tree-based models) and the evaluation metrics used. Then explain the trade-offs in performance, interpretability, and operational constraints that led you to choose logistic regression. Emphasize how you validated the decision with business stakeholders and ensured it met regulatory and explainability requirements.

Pro tip: In regulated industries like Experian, interpretability and compliance often outweigh marginal performance gains. Highlight how logistic regression's transparency and ease of explanation to regulators and clients made it the preferred choice, even if a more complex model performed slightly better.

1. List baseline models compared

Name the specific baseline models you evaluated, such as a dummy classifier, regularized logistic regression, decision trees, random forests, or gradient boosting. Mention that you used cross-validation and appropriate metrics like AUC-ROC, precision-recall, or F1-score.

2. Compare performance and trade-offs

Summarize the performance differences: e.g., tree-based models may have higher AUC but lower interpretability. Discuss trade-offs in training time, inference latency, and ease of deployment.

3. Explain decision drivers

Articulate the key factors that drove the decision: interpretability, regulatory compliance, explainability to stakeholders, robustness to overfitting, and alignment with business goals. Mention any constraints like model risk management policies.

4. Validate with stakeholders

Describe how you involved business partners, compliance, and engineering to ensure the model met their needs. Highlight any A/B tests or shadow deployments that confirmed the choice.

5. Conclude with impact

End by stating the outcome: logistic regression delivered sufficient performance with superior explainability, leading to faster adoption and regulatory approval. Mention any monitoring or retraining plans.

Key Points to Mention

  • Specific baseline models (e.g., dummy classifier, regularized logistic regression, decision trees, random forests, XGBoost)
  • Evaluation metrics (AUC-ROC, precision-recall, F1-score, calibration)
  • Interpretability and explainability (coefficients, odds ratios, SHAP values for complex models)
  • Regulatory and compliance requirements (e.g., adverse action reasons, model risk management)
  • Operational constraints (latency, scalability, maintenance)
  • Business impact and stakeholder alignment (e.g., ease of explanation to clients, faster deployment)

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

Q4

What is knowledge-informed machine learning and can you give a concrete example of it?

Technical Trade-offsData Modeling
Author's notes

I blanked for a second.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Define knowledge-informed machine learning as the integration of domain knowledge into ML models, then provide a concrete example from your experience or a well-known case. Emphasize how this approach improves model performance, interpretability, and alignment with business constraints, especially in regulated industries like credit scoring.

Pro tip: Relate the example to Experian's business context, such as credit risk modeling, to show you understand their domain and can apply the concept to real-world problems.

1. Define the concept

Clearly explain that knowledge-informed ML combines data-driven learning with domain expertise, using techniques like constraints, priors, or feature engineering.

2. Explain why it matters

Highlight benefits such as improved accuracy with limited data, better interpretability, and compliance with regulations.

3. Provide a concrete example

Describe a specific case, such as using monotonic constraints in credit scoring to ensure that higher income never decreases creditworthiness.

4. Connect to the role

Tie the example to the Data Scientist role at Experian, emphasizing how knowledge-informed ML can enhance their products and services.

Key Points to Mention

  • Definition: integrating domain knowledge into ML models
  • Techniques: Bayesian priors, monotonic constraints, feature engineering
  • Example: credit scoring with monotonic constraints
  • Benefits: improved accuracy, interpretability, regulatory compliance
  • Relevance to Experian: credit risk, fraud detection, marketing
  • Trade-offs: balancing data-driven insights with expert knowledge

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

Q5

When would you adjust the classification threshold to change your false positive or true positive rate, and is it possible to improve both at the same time by moving the threshold?

Product Analytics & MetricsTechnical Trade-offs
Author's notes

Short answer to the last part: no, moving a single threshold is a tradeoff.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by explaining the trade-off between false positives and true positives when adjusting the classification threshold, and clarify that moving the threshold alone cannot improve both simultaneously. Then, discuss scenarios where you would adjust the threshold based on business costs and benefits, and finally, mention techniques to improve both metrics by enhancing the model itself.

Pro tip: Quantify the business impact of false positives versus false negatives to justify threshold adjustments, and emphasize that model improvement, not threshold tuning, is the key to lifting both metrics simultaneously.

1. Explain the threshold trade-off

Describe how lowering the threshold increases true positives but also false positives, while raising it reduces false positives but also true positives. This sets the foundation that moving the threshold alone cannot improve both.

2. Identify when to adjust the threshold

Discuss scenarios where business costs dictate the optimal threshold, such as when false positives are very costly (e.g., fraud alerts) or when missing a positive is very costly (e.g., disease detection).

3. Clarify the impossibility of improving both via threshold

State explicitly that for a fixed model, moving the threshold traces out the ROC curve, so improving one metric necessarily worsens the other. Both cannot be improved simultaneously by threshold adjustment alone.

4. Describe how to improve both metrics

Explain that improving both requires enhancing the model's discriminative power, e.g., by adding better features, using a more complex model, or collecting more data, which shifts the entire ROC curve upward.

5. Connect to business context

Tie the discussion back to the company's context (e.g., Experian's credit risk models) by emphasizing that threshold selection should align with business objectives and cost matrices.

Key Points to Mention

  • Trade-off between false positives and true positives (or false negatives)
  • Business costs and benefits of different error types
  • ROC curve and AUC as measures of model performance
  • Threshold selection based on cost-sensitive learning
  • Model improvement techniques (feature engineering, algorithm selection, more data)
  • Precision-recall trade-off and its relevance for imbalanced datasets

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