The Ten-Minute Data Debrief
Why this prompt matters
Teams that eyeball spreadsheets live in meetings routinely turn noise into a confident story — a one-week blip becomes "the mobile funnel is broken," a bot-traffic spike becomes "engagement is up," and a correlation becomes a causal claim that gets repeated in three more meetings before anyone checks it. Walking back a wrong claim made in front of a VP costs more credibility than saying "I don't know yet" would have cost in the room — and the follow-up work to fix the wrong decision often takes days, not minutes.
What we use it for
A growth marketing manager gets a messy CSV export of last week's checkout funnel dropped in their inbox nine minutes before the Monday revenue standup with the VP of Growth — no time to build a deck, no time to double-check anything, but still expected to speak to the numbers with confidence.
Prompt
Role: Act as a senior data analyst who specializes in turning raw, unreviewed data into fast, defensible verbal briefings for stakeholders under time pressure. You are calm under deadline, precise about what the data does and does not show, and allergic to false confidence.
Context: I have [NUMBER OF MINUTES] before [MEETING TYPE AND AUDIENCE — e.g., "a weekly revenue standup with my VP of Growth"]. I have not had time to properly review the data below. My role is [YOUR ROLE/TEAM]. This data relates to [BUSINESS CONTEXT OR METRIC — e.g., "last week's checkout funnel performance by device"].
Here is the raw data: [PASTE YOUR RAW DATA HERE — CSV, table, or pasted numbers]
Task: Prepare me for this meeting in four parts:
1. The TOP 3 INSIGHTS from this data, ranked by business impact, each as a single sentence with the specific number that backs it up.
2. Up to 2 WATCH-OUTS — anomalies, outliers, or numbers that look wrong or unexplained — flagged with a one-line caveat about what else could explain them besides the obvious story.
3. 3 ANTICIPATED QUESTIONS a sharp stakeholder would ask about this data, each with a one-line answer I can say out loud without hedging.
4. DATA GAPS — anything I should openly say "I don't have that yet" about rather than guess.
Constraints:
- Do not invent numbers, trends, or context that are not in the data I pasted.
- Do not state a causal claim ("X caused Y") unless the data includes a controlled comparison or a documented change that lines up with the timing. Otherwise phrase it as a correlation and say so explicitly.
- Each insight must be sayable out loud in under 15 seconds — cut anything that needs a chart to understand.
- If the data is too thin or ambiguous to support 3 confident insights, say so instead of padding with weak ones.
- Prioritize by [WHAT MATTERS MOST TO YOUR AUDIENCE — e.g., "revenue impact" or "customer churn risk" or "deadline risk"].
Output Format:
TOP 3 INSIGHTS
1. ...
2. ...
3. ...
WATCH-OUTS
1. ...
2. ...
ANTICIPATED QUESTIONS & ANSWERS
Q1: ... / A1: ...
Q2: ... / A2: ...
Q3: ... / A3: ...
DATA GAPS TO FLAG
- ...Result
Sample input: a pasted CSV of last week's checkout funnel by device (desktop, mobile web, mobile app), 9 minutes before a Monday revenue standup.
TOP 3 INSIGHTS
- Mobile app checkout conversion dropped from 61% to 47% week-over-week, the steepest decline of any channel.
- Desktop remains the strongest channel at 68% conversion, up 2 points, and now accounts for 44% of total completed orders despite only 31% of sessions.
- Cart abandonment on mobile web held flat at 72%, suggesting last week's banner-ad test did not move behavior either way.
WATCH-OUTS
- The mobile app drop happened the same week as a payment-provider SDK update on Wednesday — timing lines up, but the data alone doesn't prove the SDK caused it; could also be the new shipping-fee display that launched Tuesday.
- Tuesday shows a session-count spike (+38%) with no matching revenue spike — worth checking if this is a bot/crawler traffic anomaly before reading anything into the conversion numbers for that day specifically.
ANTICIPATED QUESTIONS & ANSWERS
Q1: Is the mobile app drop a bug or a trend? A1: Too early to call a trend — it's one week, and it coincides with two shipped changes, so I'd want to see Monday's numbers before we conclude anything.
Q2: Should we pause the shipping-fee display test? A2: I'd hold off pausing until we isolate it from the SDK update — pulling both at once would tell us nothing.
Q3: Is desktop's growth real or just mobile's loss showing up there? A3: Desktop's absolute order count also grew, not just its share, so this looks like real growth, not just a shift.
DATA GAPS TO FLAG
- No breakdown by new vs. returning customers, so I can't say whether the mobile app drop is hitting first-time buyers harder.
- Tuesday's traffic spike needs a bot-filter check before anyone treats that day's numbers as representative.
Someone hands you a CSV nine minutes before a meeting and expects you to speak to it like you have had it for a week. This is one of the most common small emergencies in any data-adjacent job, and most people respond by either skimming for the first number that looks interesting, or opening a chart tool and hoping something jumps out before the clock runs out. Neither approach produces a briefing — it produces a guess dressed up as analysis.
Why this needs a structured prompt, not a quick summary
Asking an AI to “summarize this data” produces exactly what you'd expect: a flat paraphrase of the numbers, with no sense of what actually matters to the person you're about to talk to, and no warning about the parts of the data that are shaky. The prompt above is built around four forcing functions instead of one generic summary.
Why insights are capped at three and ranked
A real verbal briefing has room for three points before people stop listening. Capping the output at three, ranked by business impact, forces the model to make an editorial decision instead of listing everything it notices. The constraint that each insight must be “sayable out loud in under 15 seconds” exists for the same reason — it filters out anything that only makes sense with a chart in front of it.
Why watch-outs get a built-in alternative explanation
The most dangerous moment in a data briefing is the confident causal claim that turns out to be coincidence. The Watch-Outs section doesn't just flag anomalies — it requires the model to name a plausible alternative explanation for each one. That one sentence (“could also be the new shipping-fee display” in the example) is often the difference between a defensible comment and a claim you have to retract in the follow-up meeting.
Why anticipated questions matter more than the insights themselves
Stakeholders rarely argue with the headline number — they ask the second-order question that reveals whether you actually understand it. Pre-writing three likely questions with honest, non-hedging answers means you walk in already knowing where the conversation will go, instead of improvising under pressure.
Why the prompt forces you to admit what you don't know
The Data Gaps section is the part most people skip when they're nervous, and it's the part that protects you most. Saying “I don't have that broken out yet” in the room costs almost nothing. Guessing at it and being wrong costs a correction, a credibility hit, and sometimes a decision made on bad information.
How to adapt it
Swap in your actual time budget, your audience, and what that audience cares about most — revenue, churn risk, or a deadline. The structure stays the same whether the data is a funnel export, a support-ticket dump, or a weekly ops report; only the bracketed fields change.