CrackML by @ml.with.umang
Interview questions / ML System Design
ML System Design interview question

Cold Start: New Users vs New Items

Compare the system strategy for new-user and new-item cold start in a recommender.

mediumsystem designEvidence 41/1001 source reportLinkedIn

The 60-second answer

New users lack behavioral history, so use onboarding, context/session behavior, popularity priors, and fast adaptation from early interactions. New items lack engagement evidence, so use content/metadata/creator embeddings and allocate exploration traffic until behavioral signals mature.

Build the answer in this order

1
Frame the problem

New users lack behavioral history, so use onboarding, context/session behavior, popularity priors, and fast adaptation from early interactions.

2
Design the data path

New items lack engagement evidence, so use content/metadata/creator embeddings and allocate exploration traffic until behavioral signals mature.

3
Choose the modeling stack

Keep candidate generation heterogeneous so cold-start entities are not excluded before ranking.

4
Serve, evaluate, iterate

Evaluate warm vs cold slices separately and track how quickly performance converges as interaction history accumulates.

A useful interview mental model

This is the shape of a strong answer—not a script to memorize.

01Requirements
02Data
03Model / Retrieval
04Serving
05Monitor

Senior-level signal

  • Senior answers discuss exploration budgets, propensity correction, and delayed feedback.
  • Design an explicit migration path from priors/content features to collaborative representations.

What the interviewer is really testing

Product framing, data design, modeling choices, serving constraints, reliability, evaluation, and explicit trade-offs.

Likely follow-up questions

What changes at 10× traffic or data volume?
Which failure mode would you monitor first in production?
How would you evaluate this offline and online before rollout?

Common weak-answer patterns

  • Jumping to a model before defining the product contract.
  • Listing components without bottlenecks, metrics, or failure handling.
  • Ignoring data quality, serving latency, monitoring, and iteration.