Interview Drills — Stakeholder Storytelling / Presentation

Open-ended interview questions for the stakeholder/presentation round. Each has a Framework (the structure a strong answer follows), a Model answer (a concise example), and a Rubric (what an interviewer listens for). Practice thinking aloud and always lead with the answer, tie evidence to a business number, and end on a decision. The app can role-play these as mock interviews (see mock-interview.md).

The universal readout structure (answer-first): Headline the answer in one sentence → show the 2–3 numbers/charts that prove it → make the recommendation → be honest about uncertainty → keep methodology for "details on demand." Use it on almost any "present your finding" or "what would you tell the VP" question.

D1

difficulty: easy
concept: telling-stories You spent a week analyzing why sign-ups dropped. You get five minutes with the VP of Growth. How do you open?
Framework: Lead with the headline answer in one sentence (what + why) → name the 1–2 numbers that size it → state your recommendation → offer methodology only on demand. Resist the urge to walk through your process first.
Model answer: "I'd open: 'Sign-ups fell 15% last month, and it's almost entirely the new mobile checkout — completion on mobile dropped from 60% to 38%, while desktop was flat. My recommendation is to roll the mobile checkout back to the old flow this week while we fix it. Happy to walk through how I isolated it, but that's the headline.' I lead with the answer because a busy VP wants 'so what' first, not my SQL."
Rubric: Strong answers put the conclusion in the first sentence, quantify it, and reach a recommendation fast; they explicitly hold methodology for later. Weak answers start with "So first I pulled the data and cleaned it…", bury the finding, or never make a recommendation.

D2

difficulty: easy
concept: visualization An executive asks for a chart of revenue this quarter vs. last. Your tool defaults to a bar chart with the y-axis starting at $480k so the difference looks dramatic. What do you do, and why?
Framework: Name the honesty problem (truncated axis) → state the rule for the chart type → pick the right chart for the question → explain the trade-off between "looks impressive" and "is trustworthy."
Model answer: "I'd fix the axis to start at zero. A bar chart compares magnitudes by bar length, so a truncated axis exaggerates a small change — here a 4% lift would look like a doubling. If the real story is the trend across quarters I'd use a line chart instead, where a non-zero axis is acceptable. I'd rather the exec trust every chart I show than be wowed once and doubt me later; credibility compounds."
Rubric: Strong answers catch the zero-baseline issue, tie the rule to why (bars encode length), and weigh honesty over impressiveness. Weak answers keep the dramatic version to "make the point," or can't explain why a non-zero bar axis misleads.

D3

difficulty: medium
concept: telling-stories You found a bug causing a 2% checkout error rate. How do you present its severity so a non-technical leadership team takes it seriously?
Framework: Translate the technical metric into a business number (money/customers/time) → show the simplest evidence → state the recommended action and its urgency → be honest about how you estimated, so the number is defensible.
Model answer: "I wouldn't say 'error rate is 2%' — that sounds tiny. I'd say: 'This bug fails about 2% of checkouts. At our volume that's roughly 4,000 abandoned orders a month, about $60k in lost revenue, and it hits returning customers hardest.' Evidence: one bar of failed-vs-successful checkouts by week showing it started after the release. Recommendation: prioritize the fix this sprint. I'd note the $60k assumes average order value holds, so treat it as a close estimate, not a precise figure."
Rubric: Strong answers convert the percentage into dollars/customers in the audience's terms, show one clean piece of evidence, and flag the assumption behind the estimate. Weak answers leave it as "2% error rate," over-precise the dollar figure as if exact, or never connect to a decision.

D4

difficulty: medium
concept: telling-stories You're partway through your readout and a VP cuts in: "Okay, so what? What do you actually want us to do?" How do you respond?
Framework: Don't get defensive — recognize you buried the lede → give the one-line answer and the single recommended action immediately → attach the number that justifies it → offer to go deeper only if they want.
Model answer: "I'd take it as a signal I led with too much process. I'd say: 'Fair — the action is: pause the email discount for new users. It's driving sign-ups but those users churn within 30 days, so we're paying to acquire people who don't stick. Net, it's costing more than it returns by about $15k a month. I have the cohort retention curves if you want the detail.' Then I'd stop talking and let them decide."
Rubric: Strong answers stay composed, immediately surface a concrete action + supporting number, and offer detail rather than forcing it. Weak answers get flustered, retreat further into methodology, or give an analysis with no recommendation the VP can act on.

D5

difficulty: medium
concept: telling-stories Your finding is based on only two weeks of post-launch data. A director asks, "So can we be confident this works?" How do you handle the uncertainty without losing credibility?
Framework: Be honest about the limitation explicitly → still give a directional read (don't refuse to commit) → frame it as signal vs. proof → say what would confirm it and by when → tie to a low-regret next step.
Model answer: "I'd be straight: 'This is two weeks of data, so it's an early signal, not proof — the lift could be novelty. That said, the direction is clear and consistent across both weeks: the new flow is up 8% on completion. I'd treat it as promising and let it run two more weeks to confirm before we commit engineering to expand it. A low-risk next step now is keeping it on for the 50% already in the test.' Honesty here builds trust; overclaiming an 8% lift that later evaporates would cost me far more."
Rubric: Strong answers name the uncertainty plainly, still give a usable directional answer, distinguish signal from proof, and propose a sensible next step. Weak answers either overclaim certainty to sound impressive, or hedge so hard they give leadership nothing to decide on.

D6

difficulty: hard
concept: telling-stories You have one finding to present to two audiences this week: the finance team and the product team. How does your readout change for each?
Framework: Hold the underlying finding constant → identify what each audience optimizes for (finance: money/risk; product: users/experience) → reframe the headline, the chosen metric, and the recommended action per audience → keep the same honest evidence and plain language for both.
Model answer: "Say the finding is: a confusing returns flow drives support tickets and refunds. To finance I'd headline the cost: 'Returns confusion is costing ~$40k/month in refunds and support load; fixing the flow likely recovers most of it' — metric is dollars, recommendation framed as savings. To product I'd headline the experience: 'The returns step is where users get stuck — it generates 30% of our support tickets and tanks post-purchase satisfaction; let's redesign it next sprint' — same root cause, but the metric is user pain and the ask is a design change. Same honest data, two framings, each in the audience's terms."
Rubric: Strong answers keep one truthful finding but genuinely re-anchor the headline, metric, and ask to what each audience cares about — without distorting the data. Weak answers give the identical generic readout to both, or change the facts (not just the framing) to please each room.