OLX / Craigslist Classifieds#

Problem statement (interviewer prompt)

Design OLX / Craigslist: a classifieds platform where individuals post ads (text + photos), buyers search by category + location, and the two parties chat in-app to arrange a deal. No integrated checkout - focus on listings, search, moderation, and trust.

flowchart LR
  U([User])
  POST[Post Ad Service]
  MOD[Moderation]
  SRCH[Search]
  CHAT[In-app chat]
  ADS[(Listings)]
  U --> POST --> MOD --> ADS
  U --> SRCH --> ADS
  U --> CHAT

    classDef client fill:#dbeafe,stroke:#1e40af,stroke-width:1px,color:#0f172a;
    classDef edge fill:#cffafe,stroke:#0e7490,stroke-width:1px,color:#0f172a;
    classDef service fill:#fef3c7,stroke:#92400e,stroke-width:1px,color:#0f172a;
    classDef datastore fill:#fee2e2,stroke:#991b1b,stroke-width:1px,color:#0f172a;
    classDef cache fill:#fed7aa,stroke:#9a3412,stroke-width:1px,color:#0f172a;
    classDef queue fill:#ede9fe,stroke:#5b21b6,stroke-width:1px,color:#0f172a;
    classDef compute fill:#d1fae5,stroke:#065f46,stroke-width:1px,color:#0f172a;
    classDef storage fill:#e5e7eb,stroke:#374151,stroke-width:1px,color:#0f172a;
    classDef external fill:#fce7f3,stroke:#9d174d,stroke-width:1px,color:#0f172a;
    classDef obs fill:#f3e8ff,stroke:#6b21a8,stroke-width:1px,color:#0f172a;
    class U client;
    class POST,MOD,SRCH,CHAT service;
    class ADS datastore;

flowchart TB
  subgraph Apps
    BR
    MOB
  end

  subgraph Edge
    CDN
    GW
  end

  subgraph Post[Post & Manage]
    POST[Post Ad Service]
    IMG[Image upload]
    OBJ[(S3)]
    CAT[Category / Taxonomy]
    MOD[Moderation pipeline<br/>ML + human]
    NSFW
    DUPE[Duplicate detector]
  end

  subgraph Search
    SRCH[Search API]
    IDX[(Inverted index<br/>geo + filters)]
    REL([Relevance ranker])
    SAVED[Saved searches]
    ALERT[Alert notifier]
  end

  subgraph Engage
    CHAT[Chat Service]
    WS[WebSocket]
    PUSH
    EMAIL
    REPORT[Report / abuse]
  end

  subgraph Monetize
    PROMO[Featured listing]
    PAY[Payment]
    SUB[Subscription / shops]
  end

  subgraph Trust
    BAN[Ban list]
    KYC[Light KYC]
    PHONE([Phone verify])
  end

  Apps --> CDN --> GW
  GW --> Post --> ADS[(Listings)]
  Post --> Moderation:::nope
  GW --> Search --> ADS
  Search --> SAVED --> ALERT --> Engage
  GW --> Engage --> CHAT
  CHAT --> WS
  Engage --> REPORT --> Trust
  Monetize --- GW

    classDef client fill:#dbeafe,stroke:#1e40af,stroke-width:1px,color:#0f172a;
    classDef edge fill:#cffafe,stroke:#0e7490,stroke-width:1px,color:#0f172a;
    classDef service fill:#fef3c7,stroke:#92400e,stroke-width:1px,color:#0f172a;
    classDef datastore fill:#fee2e2,stroke:#991b1b,stroke-width:1px,color:#0f172a;
    classDef cache fill:#fed7aa,stroke:#9a3412,stroke-width:1px,color:#0f172a;
    classDef queue fill:#ede9fe,stroke:#5b21b6,stroke-width:1px,color:#0f172a;
    classDef compute fill:#d1fae5,stroke:#065f46,stroke-width:1px,color:#0f172a;
    classDef storage fill:#e5e7eb,stroke:#374151,stroke-width:1px,color:#0f172a;
    classDef external fill:#fce7f3,stroke:#9d174d,stroke-width:1px,color:#0f172a;
    classDef obs fill:#f3e8ff,stroke:#6b21a8,stroke-width:1px,color:#0f172a;
    class PHONE client;
    class POST,IMG,CAT,MOD,DUPE,SRCH,SAVED,CHAT,WS,REPORT,PROMO,PAY,SUB,BAN,KYC service;
    class IDX,ADS datastore;
    class REL compute;
    class OBJ storage;
    class ALERT obs;

Moderation pipeline#

On create: ML classifies for prohibited content (weapons, drugs, fraud signals).
Image NSFW model.
Duplicate detection via perceptual hash + text similarity.
Borderline cases → human queue.

Search#

Heavy geo filter - index by city / S2 cell + categorical filters.
Saved searches with periodic alert evaluation.

Glossary & fundamentals#

Concepts referenced in this design. Each row links to its canonical page; the tag column shows whether it is a high-level (HLD) or low-level (LLD) concept.

Tag	Concept	What it is	Page
`HLD`	CDN	edge caching for static assets	cdn
`HLD`	Realtime protocols	WS / SSE / polling / gRPC streaming	realtime-protocols
`HLD`	Geo indexing	Geohash, Quadtree, S2, H3, R-tree	geo-indexing
`HLD`	Search internals	inverted index, BM25, embeddings, ANN	search-internals

Quick reference#

Functional#

Post / browse classified ads (no integrated checkout).
Geo + category search.
In-app chat between buyer & seller.
Moderation.
Optional paid promotion.

Non-functional#

Eventual consistency on listings is fine.
Heavy spam / fraud filtering.
p99 search < 500 ms.

Capacity#

100M+ active listings globally for big platforms.
High image volume; aggressive transcoding.

Schema#

listings(id, owner, category, geo, price, status, body, images[])
categories(id, parent, attrs[])
users(id, phone_verified, badges[])
chats(thread_id, listing_id, [participants])

Trade-offs#

Open marketplace = scam risk; invest in moderation > features.
No payment integration (classic Craigslist) = simpler but offline fraud risk.
Chat in-app keeps the conversation auditable for trust & safety.

Refs#

OLX / Mercari / Letgo engineering blogs.
ByteByteGo "Design a classifieds platform".

FAQ#

How does OLX search listings by category and location?#

Listings are indexed in a search engine with category, price, and geo filters. Geohash or S2 cells prune by location and category facets let users narrow down quickly.

How does OLX moderate user posted ads?#

New ads pass through automated checks for category fit, banned content, and duplicate detection. Human moderators review flagged items, and reputation signals on the poster speed up approval.

How does in app chat work between buyer and seller?#

A WebSocket or long poll gateway routes messages through a chat service backed by an inbox per user. Phone numbers are masked behind a proxy number for safety.

What is different between OLX and an ecommerce site?#

OLX has no integrated checkout or inventory. The platform's job is discovery, communication, and moderation. Payment and shipping happen offline between the buyer and seller.

How does OLX handle photo uploads efficiently?#

Photos are uploaded directly from the client to object storage with presigned URLs, then a thumbnail pipeline produces multiple sizes served via CDN.