Whiteboard Scale›Topics›Click Aggregator

Ad Click Aggregator

COMMON

Click aggregation is asked at Google, Meta, and Amazon because it is the streaming interview with dollar amounts attached: 8.6 billion clicks a day feeding dashboards that must be fresh in seconds, budget enforcement where every second of lag is priced in cash, and invoices that must be exact under audit. You will count twice (a Flink stream path for decisions, a nightly batch over the immutable raw log for money), survive the viral ad with two-stage salted aggregation, walk the three-link exactly-once chain, and separate duplicates (identity) from fraud (judgment).

  • Count twice: stream for dashboards and budgets, batch over the raw log for billing, reconciliation between
  • Survive the Super Bowl ad: two-stage salted aggregation bounds the merge at 16 msg/sec regardless of volume
  • Exactly-once as a three-link chain: event identity, atomic state+offsets, and the sink everyone forgets
GoogleMetaAmazonTikTokPinterestUber
8
Concepts
Deep dives
10
Cheat Items
Quick ref
▶

Visual Solutions

Step-by-step animated walkthroughs with capacity estimation, API design, database schema, and failure modes built in.

AnimatedWatch solutions →
📋

Cheat Sheet

Key concepts, trade-offs, and quick-reference notes for Click Aggregator. Everything you need at a glance.

Quick referenceView cheat sheet →
⚠

Anti-Patterns

Common design mistakes candidates make. Wrong approaches vs correct approaches for each trap.

8 anti-patternsLearn pitfalls →
🔥

Failure Modes

What breaks in production, how to detect it, and how to fix it. Detection metrics, mitigations, and severity ratings.

5 failure modesStudy failures →
Difficulty Ladder

Start simple. Build to staff-level.

Level 1
Junior / Basics
Core concepts, single-service design, straightforward requirements
Level 2
Mid-Level Interview
Multi-service architecture, trade-off discussions, standard scaling
Level 3
Senior / Deep Dive
Complex distributed systems, failure modes, consistency guarantees
Level 4
Staff+ / FAANG Hard
Planet-scale design, novel architectures, cross-cutting concerns
Elevator Pitch3-minute interview summary

“I would design a click aggregator for 8.6 billion clicks a day, 500K per second at peak, serving three masters on different points of the speed-certainty curve. The key move: count twice. A Flink stream path: event-time windows with 15-minute lateness, two-stage salted aggregation (a viral ad costs the merge layer 16 messages a second, not 100K), checkpoint-atomic state, idempotent upsert sink: feeds dashboards in seconds and a budget lane fresh within 1-2 seconds, because staleness times spend velocity is dollars the platform eats. A nightly batch over the immutable raw log (1.3 TB/day, about $30 of S3) recomputes truth with exhaustive dedup and mature fraud verdicts, and reconciliation pages past 0.1% divergence.”

Concepts Unlocked8 concepts in this topic

Count Twice

TRICKY

Stream for decisions, batch over the raw log for money, reconciliation between

Core Aggregation Design

Exactly-Once Counting

TRICKY

Three links: event identity, atomic state+offsets, and the sink everyone forgets

Core Aggregation Design

Windows and Watermarks

STANDARD

Event time, not processing time: watermark -> fire -> lateness -> correction -> side-output

High Level System Design

Two-Stage Aggregation

STANDARD

Salt the hot ad across 16 partials; stage two merges 16 msg/sec no matter the volume

High Level System Design

The Raw Log

STANDARD

Views are disposable, truth is not: replay, reprocess, audit, bootstrap

Database Schema

The Budget Race

TRICKY

Staleness x spend velocity = dollars at risk: fail closed, pace predictively

Replication and Fault Tolerance

Dedup vs Fraud

STANDARD

Duplicates need identity (click_id); fraud needs judgment (tiered verdicts, marks not deletes)

Monitoring and Complete System

OLAP Serving

EASY

Columnar + rollups: 30-day dashboards read 720 rows, not 43,200: or billions

What is an Ad Click Aggregator