SelectFromData

Interview prep

Pass the SQL screen.

Most analyst offers are decided in one round: the live SQL screen. An unseen schema, a ticking clock, and a question you have to solve while talking. Reading model answers doesn't train that — reps do. Here you rehearse the real thing: timed screens on schemas you've never seen, graded on the result by the same engine as the rest of the platform.

The SQL Screen

Medior4 questions · 20 min Pro

Medior screen

Streamora · subscription video

The screen that decides mid-level offers: latest-per-entity, top-N per group, the NOT IN NULL trap, month-over-month.

Senior3 questions · 25 min Pro

Senior screen

PulseBoard · product analytics

Funnel conversion, a fan-out that silently doubles revenue, and a running total — senior-screen classics.

The trap, revealed

Every question secretly tests a classic mistake. The debrief names it — and shows the wrong query most people write.

One correct solution + talk track

A model query in standard SQL, plus what a strong candidate says out loud while writing it.

Run it again, free

Screens are retakeable without limit. Your best result is saved per screen.

What screens actually test

Six question shapes we see again and again — each hiding a trap. Know the shape, name the trap out loud, and you read as experienced.

Top-N

LIMIT without ORDER BY returns arbitrary rows — and it looks right.

LIMIT and Top-N

Aggregation + HAVING

Filtering groups with WHERE errors out — knowing why is the actual question.

HAVING and the Order SQL Runs In

Latest per entity

Aggregating over the full history counts every stale row as current.

Cleaning the Noise Pro

Anti-join & the NULL trap

One NULL in the subquery and NOT IN returns nothing at all.

Anti-Joins Pro

Window deltas

Month-over-month without LAG means fragile self-joins that misalign periods.

Window Functions Pro

Fan-out joins

A duplicate key on the joined side silently doubles your total.

Aggregations That Lie Pro

Talk tracks — the verbal half

What to actually say for the questions SQL can’t grade: behavioral, modeling, stakeholder rounds. Structure stories as Situation → Task → Action → Result. Tap a question to see what a strong answer covers.