Data scientist interview: how do you handle missing data in a dataset?

Model answer

Show a structured approach rather than jumping to one trick. Start by understanding the missingness: how much is missing, in which columns, and why — is it random, or does the fact that it is missing carry information, such as a field only some users fill? That diagnosis drives the choice. Then walk through options with trade-offs. Dropping rows or columns is acceptable when the missing amount is tiny or a column is mostly empty, but it wastes data and can bias results. Imputation — filling with the mean, median, mode, or a model-based estimate — keeps the data but can distort distributions if overused. Sometimes the best move is to add a flag indicating the value was missing, so the model can use that signal. Emphasise that you validate the effect of your choice by checking model performance on held-out data, not by assuming. Mention avoiding leakage — computing imputation values from the training set only. Avoid saying "I would just fill with the mean". The interviewer wants to see that you reason about why data is missing, pick a method deliberately, then verify it.

This is a guide to structure and approach, not a script to memorise — say it in your own words.

Practice this yourself

Type your own answer below and get it scored instantly by the same AI that powers Vachix mock interviews.