TikTok·Software Engineer·Onsite - System Design / Architecture·Senior

SeniorPrefer not to say

Jun 2026

Summary

TikTok system design round for a software engineering role. The whole thing was one big data modeling question with a lot of moving parts, and I left feeling like I'd covered maybe 70% of what they were actually looking for.

Questions Asked (1)

Given a business scenario and a set of required metrics, design a full data model: define your tables, each table's grain, column names and data types, nullability, primary and foreign keys, partitioning and indexing strategy, and explain how each metric gets computed from your schema. Also discuss trade-offs like star vs. snowflake schema, denormalization decisions, slowly changing dimension handling, and data quality concerns.

Data ModelingSystem DesignTechnical Trade-offs

Author's notes

This was one question but it was basically five questions stitched together.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the business scenario and metrics, then design a star schema with a central fact table at the lowest grain, surrounded by dimensions. Walk through each table's schema, keys, partitioning, and indexing, and explain how each metric is computed. Finally, discuss trade-offs and data quality considerations.

Pro tip: Emphasize how your design supports TikTok's scale and real-time analytics needs, and proactively mention slowly changing dimensions and data quality checks to show depth.

1. Clarify Requirements and Metrics

Ask clarifying questions to understand the business scenario, required metrics, data volume, and latency needs. Confirm the grain of analysis and any known dimensions.

2. Design Fact and Dimension Tables

Define a fact table at the lowest grain (e.g., one row per event) with foreign keys to dimension tables. Specify column names, data types, nullability, and primary/foreign keys for each table.

3. Define Partitioning and Indexing

Choose partitioning (e.g., by date) and indexing strategies (e.g., bitmap indexes on dimensions, sort keys on fact) to optimize query performance for the given metrics.

4. Explain Metric Computations

For each required metric, describe the SQL-like aggregation or calculation from the schema, ensuring clarity on how joins and filters are applied.

5. Discuss Trade-offs and Data Quality

Compare star vs. snowflake schema, denormalization decisions, slowly changing dimension handling (Type 1/2/3), and data quality concerns like late-arriving data, duplicates, and null handling.

Key Points to Mention

Star schema vs. snowflake schema trade-offs: query simplicity vs. storage normalization
Grain definition: ensuring the fact table grain aligns with the lowest level of detail needed for metrics
Partitioning and indexing strategies: e.g., partitioning by date, using clustered indexes for common filters
Slowly changing dimensions: Type 2 for historical tracking, Type 1 for overwrites, and impact on fact joins
Data quality: handling late-arriving facts, deduplication, null values, and data validation checks
Denormalization: when to denormalize dimensions for performance vs. maintainability

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.