Skip to main content

The LeetCode for product managers

1,347 real PM interview questions, with expert answers.

Write your answer under time, then compare it with a worked answer to the same question.

Free to browse. No credit card required. Newest questions added September 2026.

This is the practice page

A free question with its expert answer on the left, your editor on the right. Every question on the site works this way.

#183GoogleProduct SenseMedium

183. You're a PM at YouTube and are tasked with improving thumbnail quality. How would you do it?

Context

Thumbnails are the first impression for every video. Creators often rely on manual uploads, but many thumbnails fail to convey clarity or follow best practices. YouTube can augment production with ML-powered suggestions and better creator workflows to lift CTRs and satisfaction without costing perception of authenticity.

Clarifying questions

  1. Is the focus on creator tools, viewer-facing automation, or both? Primarily creators, with optional viewer fallback.
  2. Are we allowed to auto-generate thumbnails? Yes, as long as creators can override them.

Strategy

  • Creator tooling: Provide an in-product thumbnail editor with smart cropping, highlight detection, and per-platform preview.
  • ML recommendations: Analyze past frames to suggest compelling stills, overlay text for contrast, and surface metadata like expressions or landmarks.
  • Quality guardrails: Scan thumbnails for sensitive content, ensure accessibility (contrast, text readability), and flag low-quality images for improvement.

Metrics

  • North star: Creator adoption of suggested thumbnails and resulting uplift in click-through rate (CTR).
  • Engagement: Number of thumbnails generated or adjusted through the tool, average creation time saved, and reduction in reused assets.
  • Outcome: Incremental minutes watched attributed to better thumbnails, reduction in low-performing videos due to churn (stale metadata).
  • Guardrails: Creator override rate (too many auto-suggestions), false positives on sensitive content, and creator sentiment.

Execution

  1. Pilot the thumbnail editor with top creators, instrumenting time to publish and CTA to retake thumbnails.
  2. Train an ML model on successful thumbnails across languages/genres and surface alternate candidates directly in the editor.
  3. Launch viewer-facing fallback (auto-thumbnails) only when creators don’t upload one, maintaining transparency and easy opt-out.

Summary

Improving thumbnail quality means empowering creators with tools and ML signals while tracking CTR gains and safeguarding authenticity. Measured improvements in creator efficiency and viewer clicks validate the effort.

Your answer

Clarify: which surface, which creators, what does “quality” mean to a viewer?

Goal: raise click-through without raising regret (short watch after click).

|

Drafts are saved on this device until you submit.

Premium

One plan. Every question.

Premium unlocks the questions marked with a crown and the expert answers behind them.

Monthly

Month to month, cancel any time

$3.99/month
  • Expert answers on premium questions
  • Access to every premium question
  • AI feedback on the answers you submit
  • Cancel anytime
Save 58%

Annual

Best value for a full interview loop

$20/year
$1.67 a month, billed yearly
  • Expert answers on premium questions
  • Access to every premium question
  • AI feedback on the answers you submit
  • Cancel anytime

Secure checkout via Stripe. Cancel anytime.

Questions people ask