The test, drafted

Six questions short

Lina: Monday the fifteenth. The Data Analyst test opens on the twenty-ninth, and Elena is building the paper.

Elena Novak: Four numerical reasoning questions I have. Six on data analysis I don't — the topic didn't exist until this morning.

David: How long does it take to write six good questions?

Lina: With the options and an answer key, a careful hour. Nova drafts them in half a minute — and the hour moves to reading them.

Lina: The test, drafted. Nova writes questions for the bank; a recruiter reads every one before any candidate can see it.

The test register

Lina: The test register for this round.

Lina: It belongs to the Data Analyst job,

Lina: and it runs online, on the candidate portal: open from Monday the twenty-ninth to Tuesday the thirtieth, on the paper Elena set up this morning.

David: And it already knows the paper isn't ready.

The paper

Lina: The paper: ten questions per candidate, drawn at random — four from numerical reasoning, six from data analysis.

Lina: And the note says what's wrong: data analysis has no usable questions for a draw of six.

Lina: So the paper is marked not ready, and the test can't call a single candidate until it is.

David: It won't just give everyone four questions?

Lina: No. A test that quietly shrinks is worse than one that won't start.

Draft with Nova

Lina: Draft Bank Questions with Nova, with the deal written on the form: the questions go into the bank as drafts, and nothing is drawn into a paper until a recruiter reviews it.

Lina: Topic: data analysis, nothing in the bank yet.

Lina: Difficulty: medium,

Lina: and eight of them, so she has some to spare.

David: What does Nova know about the job?

Lina: The job this paper is used for: its title, department, summary and skills. Nothing about any candidate.

The rules it writes by

Lina: Every draft is written under the same five rules.

Lina: Relevant to the job,

Lina: unambiguous — one right answer, or a clear set of right answers,

Lina: free of cultural, gender, religious or regional bias,

David: So no questions about cricket for a job in Lahore.

Lina: Nor baseball for one in London. Answerable without outside material,

Lina: and never a copy of a question already in the bank.

Eight drafts

Lina: Twenty-eight seconds later, eight drafts — and the page opens on them.

Lina: Eight waiting for review, none drawable.

Lina: Every one counted as drafted with AI,

Lina: and a mix of types: true or false, a number to work out, multiple choice and single choice.

David: And the bill for that?

Lina: About three cents.

Reading one

Lina: She reads every one. This one: which SQL join keeps every row from the left table, and fills the gaps with nulls?

Lina: Four options, and the key: LEFT JOIN.

Lina: It's a draft until she says otherwise,

Lina: and the note tells her exactly what to check: the wording and the key.

David: And it's right?

Lina: It's right, and it's this job — SQL is the first thing the advert asks for.

Approved

Lina: Mark as reviewed, and the badge changes: in the bank.

David: And the tick goes into the question's history under her name, like any other change to it.

The one she drops

Lina: This one she doesn't keep. Which of these are common data cleaning tasks?

Lina: Two of the wrong answers give themselves away: deleting the whole dataset if it looks messy,

Lina: and shuffling the rows for no reason.

David: Nobody would ever pick those.

Lina: Exactly — so it only really has three options. It isn't wrong, it's weak. She archives it.

Seven kept, one archived

Lina: By ten o'clock, all eight are read.

Lina: Seven kept: correct, about this job, and fair.

Lina: One archived, because its wrong answers pointed straight at the right ones.

David: Any she hesitated over?

Lina: The ice-cream one: ice cream sales and drownings rise together — why? It's a textbook example, and easy. She keeps it, because mistaking correlation for cause is exactly what an analyst must not do.

Nova's drafts, reviewed

Lina: The bank, filtered to Nova's drafts:

Lina: seven reviewed and drawable,

Lina: none waiting,

David: Still labelled as drafted with AI, though. The label doesn't go away when a person approves them.

The paper, ready

Lina: Back on the paper: ten questions, and the warning is gone.

Lina: Seven usable questions for a draw of six, so the candidates won't all get the same six.

Composition

Lina: And the composition says it in one line: ready — ten questions in thirty minutes, a different draw for each candidate, and a reload never reshuffles.

Three easy mistakes

Lina: Three easy mistakes with drafted questions.

Lina: Approving in bulk. Ticking eight drafts without reading them is the same as not reviewing at all.

Lina: Drafting without context. Data analysis for whom? A lab, a bank, a warehouse — the job is what makes the questions yours.

David: And asking for exactly as many as the paper draws. One archived, and the paper isn't ready again.

Half an hour

Elena Novak: Half an hour. Eight read, seven kept — and every one of them I'd put my name to.

Lina: Which is the point: it is her name on them.

Lina: Next: the interview kit — Nova drafts questions for each competency, and Farah gets a preparation kit of her own.