A URL shortener (TinyURL, bit.ly) maps a long URL to a short, unique code and back again. As a full system it is a famous HLD problem — billions of redirects, global low-latency reads, sharded storage. This topic is deliberately the LLD slice: the class model and, above all, the encoding strategy that turns an ID into a short code. Scaling, caching, and geo-distribution are deferred to the HLD module.
That focus is exactly what makes it a good class-design exercise. The interesting decision — how do you generate the short code? — has three textbook answers (counter + base62, random/hash, or a hybrid), each with a different collision profile, so it maps cleanly onto the Strategy pattern.
What the interviewer is testing
- Can you separate the three concerns: encoding (id → code), storage (the repository), and orchestration (the service)?
- Do you understand why base62 over an incrementing counter is collision-free, while hashing the URL is not — and how you'd handle the collision when it happens?
- Can you keep the design extensible so a new encoding scheme (or custom aliases) drops in without rewriting the service?
How to frame it
Start by clarifying scope: read-heavy (redirects ≫ creations), short-code length and alphabet, whether custom aliases and expiration are in scope, and that horizontal scaling is an HLD concern you'll name but not design here.
Then drive the flow: shorten(longUrl) → IDGenerator produces a unique id → EncoderStrategy turns it into a code → repository stores code ↔ longUrl → return code. Resolution is the reverse: resolve(code) → repository lookup → longUrl (or 404/expired). Put encoding behind an EncoderStrategy interface, make the URLShortener service a Singleton (or a Spring bean), and use a small Factory to select the strategy by configuration.
The senior move is to be precise about collisions: a counter-driven base62 encoder is unique by construction (the id is unique), whereas a hash/random encoder needs a uniqueness check and a retry-with-salt loop. Saying that out loud separates you from candidates who hash and hope.