7-Day Crash Plan

Updated Nov 16, 2026 · 3 min read
On this page

    The recruiter emails on Monday: your system design round is next Monday. You have a full-time job, about an hour and a half each evening, and a list of 50 topics you feel you should know. If you try to cover them all, you'll arrive having skimmed everything and practised nothing.

    A week is enough to get noticeably better, but only if you spend it on the things that move the score: a reliable framework, the core building blocks, and a few full problems practised out loud. This plan does exactly that, one focus per day.

    How it works

    Seven days, one skill per day: framework, building blocks, URL shortener, rate limiter, feed or chat, booking, and a timed mock with review.
    Seven days, one skill per day: framework, building blocks, URL shortener, rate limiter, feed or chat, booking, and a timed mock with review.

    Day 1: How the round works

    Day 2: Core building blocks

    Days 3–6: One problem a day

    Use the same session every day:

    One 75-minute practice session: 5 minutes reading the prompt, 25 minutes designing on paper with a timer, 20 minutes reading the walkthrough, 15 minutes fixing gaps, and 10 minutes saying the design o
    One 75-minute practice session: 5 minutes reading the prompt, 25 minutes designing on paper with a timer, 20 minutes reading the walkthrough, 15 minutes fixing gaps, and 10 minutes saying the design out loud.
    DayProblemWhat it trains
    3Design a URL ShortenerThe full framework on a simple system: IDs, read-heavy caching
    4Design a Rate LimiterA focused deep dive: algorithms, shared counters, failure policy
    5Design FB News Feed or Design a Chat SystemFanout or real-time delivery, the two most common "scale" themes
    6Design a Ticket Booking SystemCorrectness under contention: holds, locks, idempotency

    When a walkthrough uses a toolkit topic you don't know (for example Idempotency on Day 6), read that article in the "fix your gaps" block.

    Day 7: Mock and review

    Trade-offs and costs

    • You will not cover everything. Sharding, consistency models and stream processing get only what the four problems bring up. That is the right trade for one week.
    • Depth over breadth: four problems done properly beat twelve skimmed.
    • Fatigue: 90 minutes after work is enough. A tired, rushed session builds less than a short focused one.

    When to reach for it

    • Interview in 5–10 days, with 60–90 minutes a day.
    • Three days only: Day 1 (framework), Day 3 (URL shortener), Day 7 (mock). Read the building blocks inside the URL shortener walkthrough as they come up.
    • Two weeks or more: use the 30-Day Study Plan and compress it.
    • Already experienced: skip Day 2, replace Day 3 with a harder problem, and do two mocks.

    Example

    An engineer's Day 5 with the news feed:

    • 0–5 min: read the prompt; write questions: how many users, follows per user, celebrity accounts, ranked or chronological?
    • 5–30 min: on paper: requirements with numbers (200 million daily users, 20 reads per post), API (POST /posts, GET /feed), a design with a Post Service, Feed Service and a feed cache, plus a note: "what about celebrities?"
    • 30–50 min: read the walkthrough. Missed: the hybrid push/pull approach and capping feed lists.
    • 50–65 min: read Fanout (Read vs Write).
    • 65–75 min: explain the whole design out loud in 5 minutes, then the fanout trade-off in 2 minutes.

    Common mistakes

    • Reading all week and doing one mock at the end. Practise out loud every day from Day 3.
    • Skipping the 25-minute attempt and going straight to the walkthrough.
    • Picking only problems you already know.
    • Learning a brand-new topic the night before. Review instead.
    • Ignoring sleep before the interview. A rested brain explains better.

    Say this in the interview

    After a week of practice, your opening should be automatic:

    "Let me confirm the core use cases and scale first. I'll state assumptions as numbers, get a simple design working end to end, and then spend most of our time on the two or three riskiest parts."

    Used in