Practice/Meta/Design Catalog System

Design Catalog System

Product DesignMust

Problem Statement

You are tasked with designing a marketplace listing system that enables sellers to create product listings by uploading item descriptions, pricing information, and multiple images. The system must automatically screen submissions to detect and block prohibited content such as weapons, counterfeit goods, and other policy violations. Users should receive immediate feedback on whether their listing was accepted or rejected, along with specific reasons for any rejections.

The platform is expected to handle millions of active sellers with hundreds of thousands of new listings created daily. Peak traffic occurs during holiday seasons and promotional events, where submission rates can triple. The system must balance speed of approval with accuracy of content moderation, as false positives frustrate legitimate sellers while false negatives expose the marketplace to legal and reputational risks.

Key Requirements

Functional

Listing Creation -- sellers can submit products with title, description, price, category, and 5-10 images
Automated Content Moderation -- system analyzes text and images to detect prohibited items, offensive content, and policy violations
Real-time Feedback -- sellers receive approval or rejection status within seconds of submission
Multi-stage Review -- rejected items can be appealed for human review
Search and Discovery -- approved listings appear in search results and category browsing within minutes

Non-Functional

Scalability -- support 100K+ listing submissions per hour during peak times
Reliability -- 99.9% uptime for submission pipeline; no data loss on accepted listings
Latency -- automated moderation completes within 3-5 seconds; listing appears in search within 2 minutes
Consistency -- eventual consistency acceptable for search indexing; strong consistency for moderation decisions

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Content Moderation Pipeline Design

The moderation system is the most critical and complex component. Interviewers want to see how you balance accuracy, speed, and cost while handling diverse content types.

Hints to consider:

Consider a multi-tiered approach combining rule-based filters, ML models, and human review
Discuss how you would handle image analysis separately from text analysis given different processing times
Think about queueing strategies to handle traffic spikes without degrading moderation quality
Address the feedback loop for continuously improving moderation models based on appeal outcomes

2. Storage Architecture for Media Assets

Image handling at scale involves significant storage, bandwidth, and processing considerations that reveal architectural thinking.

Hints to consider:

Consider object storage solutions for raw uploads versus CDN-optimized versions
Discuss image preprocessing pipeline for generating thumbnails and multiple resolutions
Think about how to prevent duplicate uploads and optimize storage costs
Address backup and disaster recovery for user-uploaded content

3. Handling Asynchronous Processing and User Feedback

The system involves multiple asynchronous steps while maintaining user experience expectations around feedback timing.

Hints to consider:

Design a state machine for listing status transitions from submission through moderation to publication
Consider websockets, polling, or push notifications for real-time status updates
Think about retry mechanisms when downstream services fail during processing
Address idempotency to handle duplicate submissions from network retries

4. Search Indexing and Consistency Trade-offs

Making listings immediately searchable after approval while maintaining system scalability requires careful consistency decisions.

Hints to consider:

Discuss event-driven architecture to trigger search indexing upon approval
Consider whether to use synchronous indexing with increased latency or asynchronous with temporary inconsistency
Think about cache invalidation strategies for category pages and search results
Address how to handle listings that are later removed for policy violations

Practice/Meta/Design Catalog System

Design Catalog System

Product DesignMust

Problem Statement

Key Requirements

Functional

Listing Creation -- sellers can submit products with title, description, price, category, and 5-10 images
Automated Content Moderation -- system analyzes text and images to detect prohibited items, offensive content, and policy violations
Real-time Feedback -- sellers receive approval or rejection status within seconds of submission
Multi-stage Review -- rejected items can be appealed for human review
Search and Discovery -- approved listings appear in search results and category browsing within minutes

Non-Functional

Scalability -- support 100K+ listing submissions per hour during peak times
Reliability -- 99.9% uptime for submission pipeline; no data loss on accepted listings
Latency -- automated moderation completes within 3-5 seconds; listing appears in search within 2 minutes
Consistency -- eventual consistency acceptable for search indexing; strong consistency for moderation decisions

What Interviewers Focus On

Based on real interview experiences, these are the areas interviewers probe most deeply:

1. Content Moderation Pipeline Design

The moderation system is the most critical and complex component. Interviewers want to see how you balance accuracy, speed, and cost while handling diverse content types.

Hints to consider:

Consider a multi-tiered approach combining rule-based filters, ML models, and human review
Discuss how you would handle image analysis separately from text analysis given different processing times
Think about queueing strategies to handle traffic spikes without degrading moderation quality
Address the feedback loop for continuously improving moderation models based on appeal outcomes

2. Storage Architecture for Media Assets

Image handling at scale involves significant storage, bandwidth, and processing considerations that reveal architectural thinking.

Hints to consider:

Consider object storage solutions for raw uploads versus CDN-optimized versions
Discuss image preprocessing pipeline for generating thumbnails and multiple resolutions
Think about how to prevent duplicate uploads and optimize storage costs
Address backup and disaster recovery for user-uploaded content

3. Handling Asynchronous Processing and User Feedback

The system involves multiple asynchronous steps while maintaining user experience expectations around feedback timing.

Hints to consider:

Design a state machine for listing status transitions from submission through moderation to publication
Consider websockets, polling, or push notifications for real-time status updates
Think about retry mechanisms when downstream services fail during processing
Address idempotency to handle duplicate submissions from network retries

4. Search Indexing and Consistency Trade-offs

Making listings immediately searchable after approval while maintaining system scalability requires careful consistency decisions.

Hints to consider:

Discuss event-driven architecture to trigger search indexing upon approval
Consider whether to use synchronous indexing with increased latency or asynchronous with temporary inconsistency
Think about cache invalidation strategies for category pages and search results
Address how to handle listings that are later removed for policy violations