OneMain Financial Interview Questions — Page 1

OneMain Financial·Data Scientist·Technical Phone Screen

Apr 2026

Technical screen for a Data Scientist role at OneMain Financial. The whole thing was basically one long stats and experimentation question broken into four parts. Pretty rigorous for a phone screen, way more math-heavy than I expected.

For a two-proportion z-test with baseline conversion of 10% and a target uplift of +1 percentage point, derive the required sample size per arm. Show your formula and actual numbers.
You are also tracking 10 secondary metrics and need to control the family-wise error rate. What is the adjusted per-metric significance level under Bonferroni correction, and how does that affect your required sample size?
If you plan to peek at results daily over 14 days, describe a valid sequential testing approach such as O'Brien-Fleming or Pocock alpha-spending and explain how it changes the stopping boundary.
What is the practical cost of a Type II error in this specific test, and give one concrete way to reduce beta without increasing alpha. Quantify the trade-off.

“I knew the formula going in but fumbled the plug-in.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

Apr 2026

Technical screen for a Data Scientist role at OneMain Financial, heavy on XGBoost internals and imbalanced classification. The questions were genuinely hard and felt like they were written by someone who actually uses this stuff in production, not pulled from a generic ML interview bank.

You have a large binary classification dataset with severe class imbalance (1% positive rate), 100 features, and a hard 5-minute training time limit on modest hardware. Walk through your initial XGBoost hyperparameter choices and justify each one in terms of bias-variance tradeoff, class imbalance, and compute constraints.
Describe an efficient hyperparameter tuning strategy for this setup, including how you'd define the search space, use early stopping, and structure cross-validation to avoid data leakage when users appear across multiple folds.
Explain how XGBoost handles missing values internally during tree construction, and how that behavior interacts differently with one-hot encoded features versus target encoded features.
Compare scale_pos_weight, weighted loss, and focal loss for handling severe minority-class imbalance in XGBoost. When would you actually choose one over the others?

“This is where I spent most of my mental energy.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

May 2026

Interviewed for a Data Scientist role at OneMain Financial and got a meaty case study centered on credit card product optimization. It was more of a business strategy exercise than a pure data science problem, which I wasn't fully expecting.

You're given cost, revenue, and risk figures for a new credit card product. How would you determine the optimal pricing, credit limit, and target customer segment? And what metrics would you track after launch to evaluate whether the strategy is working?

“This one took me a second to orient.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

May 2026

Interviewed for a Data Scientist role at OneMain Financial and got a pretty focused technical screen centered entirely on decision trees and ensemble methods. Nothing too surprising but the depth they expected on the math behind splitting criteria was more than I anticipated.

Walk me through how a decision tree works for classification and regression tasks.
How does a decision tree decide where to split, and what objective functions are used for classification versus regression?
What are the main hyperparameters of a decision tree and how do they influence the bias-variance tradeoff?
Pick an ensemble method that builds on decision trees, explain how it works, and tell me when you'd choose it over a single tree.

“I covered the basics fine, recursive splitting, leaf node predictions, stopping conditions.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

May 2026

Technical screen for a Data Scientist role at OneMain Financial where they walked through a past modeling project and then pivoted into stats fundamentals. Pretty conversational but they definitely wanted to see if you actually understood the theory behind the work you claimed to have done.

What are Type I and Type II errors in hypothesis testing? Walk through the difference and give a real-world example of each. Then explain how you'd think about balancing the two in an A/B test.

“I started with the textbook definitions, false positive vs false negative, which was fine.”

View Post

OneMain Financial·Data Scientist·Onsite - Cross-functional / Panel

May 2026

Panel presentation round for a Data Scientist role at OneMain Financial. You pick a past project and walk through the whole thing in front of four or five people, which is more nerve-wracking than it sounds.

Walk the panel through a past analytics project from start to finish, covering the problem, data, methodology, results, and business impact. Also explain how you managed stakeholder expectations and handled feedback along the way.

“The part I underestimated was the stakeholder piece.”

View Post

OneMain Financial·Data Scientist·Onsite - System Design / Architecture

May 2026

System design round at OneMain Financial for a Data Scientist role. The main question was a full LRU cache design with a lot of follow-ups layered on top, more than I expected for a DS position.

Design and implement an LRU cache that supports get and put operations in O(1) average time with a fixed capacity N. Walk through your choice of data structures and how you handle updates to existing keys.
What exactly happens when a put operation exceeds the cache capacity? Define the eviction behavior precisely.
How would you make this cache thread-safe, and what are the tradeoffs involved?
How would you add an optional per-key TTL to this cache without breaking the O(1) time complexity guarantees?
What is the time and space complexity of your implementation, and what edge cases should be handled, such as a capacity of 1, repeated gets on the same key, or very large values?

“I went straight to hashmap plus doubly linked list, which is the right call, but I fumbled explaining why the doubly linked list specifically.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

May 2026

Interview for a Data Scientist role at OneMain Financial, focused entirely on walking through a data science project end-to-end. One long, structured question that basically became the whole conversation.

Walk me through a data science or analytics project you owned from start to finish, covering the problem, data, methodology, evaluation, and impact.

“This is the kind of question that sounds easy until you're actually in it and realize you've been rambling about data cleaning for three minutes.”

View Post

OneMain Financial·Data Scientist·Technical Phone Screen

Jun 2026

Interviewed for a Data Scientist role at OneMain Financial and got a probability/optimization puzzle that felt deceptively simple at first glance.

You have 100 red balls and 100 blue balls to distribute across two baskets however you like. One ball is drawn at random from each basket. How do you split the balls to maximize the probability that both draws come up red, and what is that maximum probability?

“I stared at this for a second because my instinct was to split evenly, like 50 red and 50 blue in each basket, and then I realized that gives you 50/100 times 50/100 which is just 25%.”

View Post

OneMain Financial·Data Scientist·Onsite - Product Sense / Strategy

Jun 2026

Case-style interview for a Data Scientist role at OneMain Financial, centered entirely on a credit card acquisition business case comparing branch vs. digital channels. Five questions total, ranging from strategic framing to unit economics to fraud risk. Pretty intense for a DS interview, felt more like a product strategy or biz ops loop.

Why would OneMain want to pull back from branch-based credit card acquisition and invest more in digital instead? Walk through your reasoning across costs, revenue, risk, customer experience, and competitive dynamics.
OneMain expects 100,000 new card customers in Year 1. Given branch contribution margin of $100 per customer, digital contribution margin of $200 per customer, and $11 million in fixed costs, what minimum fraction of customers needs to come from digital to break even?
The break-even digital share comes out to a small single-digit or low double-digit percentage. How do you interpret that result, and what assumptions might make it misleading?
Propose a concrete plan to grow digital acquisition share. Cover the customer funnel from awareness through activation, any pricing or referral levers you'd pull, and what experiments and metrics you'd run.
Does moving customers from branch to digital onboarding typically increase fraud risk, credit risk, or both? Define each, explain the mechanisms, and describe how you'd monitor and mitigate them.

“I went cost-first which felt safe, but I think the interviewer wanted me to lead with something more interesting like scalability or competitive pressure.”

View Post