DATA ARCHITECTURECapstone

Applied Data Modeling: How To Answer Any Modeling Question

How data engineers design data that lasts — pick a topic on the left and its full breakdown loads here: the mental model, ERDs and worked schemas, trade-offs, edge cases, and the decisions that separate a durable model from a fragile one.

⏱ 18 min readTopics chapter readerLevel · Specialized & Applied
01 · Orientation

The Capstone: Knowing vs Doing

Do not start at tables. Start at "what must this answer?", declare the grain, choose the shape for the workload, handle history, and narrate the tradeoffs.

⏱ 4 min · Topic 1 of 10

Twenty-one chapters gave you normalization, keys, grain, stars, SCDs, temporal, vault, NoSQL and graphs. This one answers what all of it was for: handed a vague request and a clock, how do you drive it to a defensible model and say why?

The gap is not knowledge. It is order, under pressure, on a domain you have never seen. Read both transcripts below, then find the minute candidate A’s answer actually went wrong.

Core mental model

Do not start at tables. Start at "what must this answer?", declare the grain, choose the shape for the workload, handle history, and narrate the tradeoffs.

Why it matters

Everything from here is rehearsal rather than instruction. The other chapters own the concepts; this one drills applying them in sequence, fast, out loud, and defending the result.

modeling method
A repeatable sequence: clarify → grain → entities/keys → shape → history → evolve.
first fault
The earliest decision a broken answer inherits from; everything after it is a symptom.
narration
Explaining the modeling reasoning out loud, the thing interviews actually grade.
Selkie Swim School · "Design the data model for our swim school."Two candidates, the same prompt, the same thirteen minutes.Both know normalisation, grain and SCDs. Only one of them ran a method. Read candidate A, then find the minute the answer actually went wrong — it is earlier than it looks.
find the first fault
Why this chapter existsCandidate A was not short of knowledge. Every fact needed to fix minute 05:10 and minute 08:20 is in chapters 4 and 11, and A could have recited both. What A lacked was an order to apply them in, under time pressure, on a domain never seen before. That is a separate skill and it is the only thing this chapter drills.
Common mistake

Jumping straight to tables on any modeling prompt. You bake in guesses about grain, history, and workload that are expensive to undo; method first.

Fixing the symptom the interviewer just poked at, instead of the decision that caused it. You bolt on a table that cannot join to the rows already written, and the real fault survives.

Better habit

Run the same method on every modeling question.

Clarify and declare grain before designing anything.

Narrate the reasoning and the tradeoffs, not just the schema.

The big idea

Modeling skill is a repeatable process, not a memorized set of schemas. Master the process and you can model a domain you have never seen.

This chapter is a rehearsal room

Every widget here is a drill against a clock: pick a question, pick a grain, take a curveball, defend a tradeoff. Where a drill catches you out, it names the chapter that owns the concept.

Remember this

Applied modeling is a repeatable method, clarify, grain, entities/keys, shape, history, evolve, narrated out loud; this chapter turns the whole course into that one usable workflow.