Squarepoint·Software Engineer·Technical Phone Screen·Intermediate

IntermediatePrefer not to say

Jun 2026

Summary

ML system design round at Squarepoint for a quant researcher role, focused on credit card transaction data. The first part was mostly about data cleaning basics, which felt a bit anticlimactic given the 'finance intuition' billing.

Questions Asked (1)

You're given a credit card spending time series with timestamp, customer ID, and merchant fields. How would you handle missing values in this dataset?

Data ModelingSystem DesignTechnical Trade-offs

Author's notes

Expected something more finance-flavored given how the question was framed beforehand, so the NA handling opener felt weirdly basic.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the data's structure and the purpose of the analysis, then systematically classify missingness by field and pattern. Propose tailored imputation or exclusion strategies for each field, emphasizing validation and documentation of assumptions.

Pro tip: Highlight that missing timestamps or customer IDs often indicate data pipeline issues, so fixing the source is better than imputing. For merchant, consider that missingness may be informative (e.g., declined transactions) and could be encoded as a separate category.

1. Clarify Data and Objectives

Ask about the data source, volume, and intended use (e.g., fraud detection, customer profiling). Confirm whether missingness is random or systematic.

2. Profile Missingness

Compute missing rates per field and analyze patterns (e.g., missing timestamps correlated with certain customers). Determine if missingness is MCAR, MAR, or MNAR.

3. Handle Missing Timestamps

If timestamps are missing, consider dropping records or imputing based on sequence if time order is known. For critical analysis, flag and exclude.

4. Handle Missing Customer IDs

If customer ID is missing, assess if it can be inferred from other records (e.g., same timestamp and merchant). Otherwise, treat as a separate 'unknown' category or exclude.

5. Handle Missing Merchants

Impute using mode or a model based on customer history, or create an 'unknown' category. Consider if missingness itself is predictive.

6. Validate and Document

Test the impact of imputation on downstream metrics. Document all decisions and assumptions for reproducibility.

Key Points to Mention

Distinguish between MCAR, MAR, and MNAR missingness mechanisms.
Consider the business context: missing merchant may indicate a declined transaction or data error.
Evaluate trade-offs between deletion, imputation, and flagging missing values.
Use domain knowledge to infer missing customer IDs from transaction patterns.
Ensure imputation methods are validated to avoid introducing bias.
Document all preprocessing steps for auditability and reproducibility.

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