# Design a personalized search ranking system: system design interview question

> Personalized search must balance query relevance, explicit preferences, sparse histories, recent activity, consent, and low latency.

Source: https://sysround.com/questions/design-personalized-search-ranking-system
Updated: 2026-11-05


Difficulty: hard. Levels: Senior, Staff+.

## How the interviewer opens it

"Design a search service that adapts results to a user's preferences and recent activity. Keep results relevant to the current query. Support users with little history and respect their privacy choices."

## What to clarify first

- Domain and scale
- Traffic and read:write mix
- Latency and freshness
- Query relevance
- Signals and sparse histories
- Privacy
- Consistency and availability
- Durability and event delivery
- Quality evidence
- Exclusions

## Scope of a strong answer

Core requirements:

- Return relevant search results that respect the user's filters.
- Adapt result order to consented preferences and recent activity.
- Use explicit preferences even when the user has no activity history.
- Honor consent withdrawal, retention limits, and deletion requests.

Usually out of scope:

- Build the base text search engine.
- Serve ads or recommendations outside search.
- Infer sensitive traits.

## What interviewers probe

### Query relevance and user preferences

A user's past interests can conflict with the current query. More clicks can hide worse search results.

### Preferences and sparse histories

Explicit preferences can exist before any activity. Sparse histories can reflect temporary needs instead of lasting interests.

### Consent and deletion

Old events can survive consent withdrawal and later consent renewal. Delayed work must not restore personal data without valid consent.

### Search latency

Personalization adds data access and score computation to every search. A slow activity read must not discard available preferences or delay useful results.

## Common mistakes

- The design lets past interests override an exact query or a hard filter.
- The design treats click rate as the only measure of search quality.
- The design ignores saved preferences when the user has no activity profile.
- The search path reads raw histories or makes a remote call for each candidate.
- The privacy plan deletes activity profiles but ignores preferences, logs, derived features, offline copies, and backups.
- The service trusts stale consent or assigns new consent epochs to old events.

## The bar by level

- **Senior:** Defines event deduplication, feature age, and request deadlines. Preserves available preferences during activity failures. Binds events to their original consent period. Prevents consent races and rejects stale work. Separates relevance tests from click gains.
- **Staff+:** Demonstrates deletion deadlines across offline jobs, backups, and derived data. Defines safe recovery after deletion. Addresses exposure bias and experiment validity. Justifies quality, latency, and cost choices. Cross-region consent remains an optional extension.

Practice this question with an AI interviewer: https://sysround.com/practice/new?problem=personalized-search


---
SysRound (https://sysround.com) is an AI practice tool for system design interviews. It has a voice or typed AI interviewer and a shared whiteboard. At the end, you get feedback on each skill and a model answer. The AI feedback can be wrong.