OOD & LLD Reference/The LLD Interview Process

Trade-offs, Extensibility & Communication

What interviewers score at senior level: naming extensibility points, acknowledging YAGNI, comparing two designs, and driving the conversation instead of silent coding.

3/5Overview: 25m

What senior rubrics actually score

Beyond correctness, staff-level loops evaluate whether you prioritize, compare alternatives, and communicate trade-offs without rambling. Silence while coding is a negative signal; over-explaining every line is too.

DimensionJunior signalSenior signal
Extensibility"I'd use interfaces everywhere""I'd extract PricingStrategy when tier 2 appears — not now"
Trade-offsOne design, no alternatives"HashMap is O(1) but unordered; TreeMap costs log n for sorted iteration"
ScopeBuilds unprompted featuresExplicitly lists out-of-scope items
CommunicationCode-firstRequirements → API → classes → code
DepthStops at happy pathNames failure mode + where it's handled

The extensibility conversation

When asked "how would you extend this?", point to one interface and one new class — not a framework. Example: "Adding SlackNotifier means implementing NotificationChannel and registering it — NotificationService unchanged." Pair every extension point with YAGNI: "I wouldn't build a plugin registry until we have three channels."

Comparing two designs aloud

Template: "Design A is simpler and ships faster; Design B isolates [axis of change] behind [interface]. Given the interviewer said pricing will vary soon, I'd pick B for OCP, accepting two extra classes." Always state what you're deferring (persistence, distributed locks, metrics).

Handling pushback

If the interviewer adds a requirement mid-round, trace impact: "That affects PricingStrategy, not ParkingLot — I'd add a new strategy class." Showing localized change is stronger than redrawing everything.

Senior-level signal

Record yourself for 2 minutes on any classic problem. Count "I would defer…" and "the trade-off is…" statements — aim for at least three. Interviewers remember candidates who made intentional cuts, not those who built everything.

Where this goes next

Parking Lot is the canonical allocation problem — entity modeling, Strategy for pricing, and a full happy-path walkthrough in one design.

Further Reading

Practice Tasks (Optional)

Design or implement locally in any language — no autograding. Focus on class structure, extensibility, and being able to explain trade-offs out loud.

  • Practice narrating trade-offs out loud

    For LRU cache vs simple HashMap cache: list 3 trade-offs (memory, thread-safety, eviction policy) as if explaining to an interviewer. Record 2 minutes of audio. Senior signal = you state what you'd defer.

    20m