Day-Of Tactics

The morning of

• Reread the JD once. Underline the three lines that matter most. You'll refer back to them in "why this role."
• Reread your résumé. Refresh dates and team sizes — these come up.
• Reread 01-the-roles. Have the role-specific framing fresh.
• Pull up 15-interview-questions and skim section A — the background framing is the first thing they'll test.
• Walk before the first session. 15 minutes. Don't try to cram in the last hour.
• Have water, scratch paper, pen. Close every tab you don't need.

Structural moves

The high-leverage shape-of-conversation moves that work across every round:

1. Always restate the problem before answering

Don't dive in. Say: "Let me make sure I understand. You're asking…" and play it back. Buys you 20 seconds to think and confirms the question you're actually answering.

2. State your approach before doing the work

"I'm going to think about this in three steps — first X, then Y, then Z. Does that sound right?" Lets the interviewer redirect early if your framing is off.

3. Pick one answer, then caveat it

"My recommendation would be X. The case where I'd change my mind is [specific condition]." Stronger than "it depends, here are five options" — interviewers want a committed recommendation with awareness of edge cases.

4. Narrate when stuck

Silence is the enemy. "I'm thinking about whether this is a propensity or a causal question…" — the interviewer will often nudge, and the narration is itself signal that you're thinking systematically.

5. End each round with a "what I'd do differently with more time"

Shows self-awareness. Even if the round went well, naming what a deeper version would look like is differentiating.

Traps to watch

The leading question

"Don't you think we should use propensity scoring here?" Be willing to push back if the answer is actually no. "I'd actually reach for DiD first — propensity assumes no unobserved confounding, and in this case [specific reason]." Capitulating earns less than a defensible disagreement.

The "what would you do at scale" follow-up

Always be ready for it. The right shape: "What I just described would handle ~1M rows. At 1B rows, I'd switch to [SQL pushdown / approximate algorithms / cluster-randomized / etc.]" Even if you're not certain of the scale answer, naming the dimension you'd worry about is signal.

The trick question

"This dataset has class imbalance of 99/1. What model would you use?" If they're testing whether you'll panic about imbalance — don't. Walk through the standard playbook: cost-sensitive learning, threshold tuning, precision-recall analysis. Don't reach for SMOTE first; senior practice favors threshold tuning and class weights.

The "show me your work" round

Some loops ask you to walk through a past analysis end-to-end. Have one analysis prepared with: the decision it informed, the method choice and alternatives, the limitations you called out, the recommendation, the outcome. 10-minute version and a 3-minute version.

Recovery patterns

You said something wrong and noticed

Correct it immediately. "Actually, what I just said about [X] is wrong — the correct framing is [Y]. Let me redo that piece." Self-correction is a positive signal, not a negative one.

You're stuck on a coding problem

State what you've tried, name the constraint that's blocking, ask for a hint if 5 minutes have passed. Stuck silently for 10 minutes is much worse than stuck verbally for 5.

You don't know the answer

Three-part response: name the gap precisely, name the adjacent thing you do know, name how you'd find the real answer. "I haven't used X in production. What I have used is the related Y. To answer this rigorously, I'd want to look at [specific resource] and verify." This works much better than guessing.

The interviewer disagrees with your answer

Don't fold instantly. Ask one clarifying question to understand their framing. If you still disagree, defend politely; if their framing is better, update visibly. Both moves score.

Questions to ask them

The questions you ask are part of the signal you send. Have at least three ready for each interviewer:

Universal (work in any round)

• "What does failure look like in this role at 90 days? At a year?"
• "What's the analytical decision the team has gotten wrong most recently, and what's been learned from it?"
• "What does the path from this role to the next look like?"

For a hiring manager round

• "How is success for this team measured at the company level?"
• "What's the team's split between strategic, operational, and ad-hoc work?"
• "Where does the team feel under-resourced right now?"

For a peer round

• "What's a recent analysis you're proud of?"
• "What's the working relationship like with product and eng?"
• "What would you change about how the team operates if you could?"

For a leader / exec round

• "What's the most important analytical question on your roadmap that nobody has answered yet?"
• "How does data work plug into strategy at this company — versus being downstream of it?"
• "What's a recent decision data influenced that wouldn't have happened otherwise?"

Closing statement

When they say "any final questions?" or "anything else you want us to know?", have a 30-second closing ready. Three beats:

1. One specific thing that confirmed your interest. "What you said about [X] confirmed what I'd hoped about this role."
2. One specific thing you'd bring. Not a recitation of your résumé — one concrete commitment. "If I joined, the first thing I'd want to dig into is [specific work]."
3. A clear ask. "What are the next steps and what's the timeline?"

Don't pad. Brevity here is its own signal.

After the loop

• Thank-you email within 24 hours to your recruiter (or to interviewers individually if you have their emails). One paragraph. Reference a specific thing from one round. Don't recap your résumé.
• Honest debrief with yourself: which rounds did you crush, which were rocky? Specific weaknesses go into your next study cycle.
• Don't refresh your inbox every 5 minutes. Loops typically take 3–7 business days to resolve internally; pestering the recruiter doesn't help.
• If you get a "no": ask for feedback specifically. Most won't give substantive feedback, but the framing of the rejection sometimes tells you which round it was.
• If you get a "yes": negotiate. Always. Even mid-level roles have room. Senior roles have substantial room.