Verified against ChatGPT · 2026-08-11
Turn a raw KPI spreadsheet dump into a monthly review narrative leadership can actually act on
Converts a list of KPI numbers into a monthly review narrative that explains what moved, why, and what decision it implies — instead of a flat status recap nobody remembers by the next meeting.
The prompt
Ready to copy — highlighted parts are example details you can swap.
You are turning a raw list of monthly KPIs into a short narrative for a review meeting — the goal is a story leadership can act on, not a restated spreadsheet. KPI LIST WITH TARGETS AND ACTUALS Customer churn: target 3%, actual 4.2%. Support ticket resolution time: target 24hrs, actual 19hrs. NPS: target 45, actual 41. TREND VS PRIOR MONTH Churn was 3.1% last month, so it worsened. Resolution time improved from 27hrs. NPS was flat at 42. KNOWN EXTERNAL FACTOR A pricing increase took effect mid-month, and a competitor launched a heavily discounted plan the same week. WHO THIS REVIEW IS FOR The VP of Customer Success and the CFO, neither of whom looks at the raw dashboards day-to-day. DECISION NEEDED FROM THIS REVIEW Whether to roll back the recent pricing increase before it hits next month's renewal cohort. PHASE 1 — SORT BY WHAT MATTERS, NOT BY LIST ORDER Re-rank the KPIs by how much each one actually changes the decision named above, not in the order they were given to you. A metric that's far off target but irrelevant to the decision gets a brief mention; a metric close to target but directly tied to the decision gets the most space. PHASE 2 — EXPLAIN THE MOVEMENT, DON'T JUST REPORT IT For each KPI you give real space to, state the number, whether it's better or worse than the trend, and a plausible reason tied to the known external factor where relevant — but do not invent a cause that wasn't given to you; if the reason for a move isn't clear from what I gave you, say the cause is unconfirmed rather than fabricating an explanation that sounds authoritative. PHASE 3 — CONNECT TO THE DECISION End with a direct statement of what this month's data implies for the specific decision leadership needs to make, stated as a recommendation with the confidence level attached — not a neutral 'the data suggests several possibilities' non-answer. WHAT NOT TO DO Do not describe every KPI with equal length regardless of relevance — that's what makes scorecard reviews forgettable. Do not use unexplained jargon the stated audience wouldn't already use themselves; match the register to who's in the room. OUTPUT FORMAT 1. One-paragraph headline: the single most important thing this month's numbers say. 2. Two to four KPI call-outs, ranked by relevance to the decision, each with number, direction, and cause (or 'cause unconfirmed'). 3. A one-paragraph recommendation tied explicitly to the stated decision, with a stated confidence level.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A flat KPI list gives every metric the same visual weight, which is exactly why scorecard reviews are forgettable — GPT-5.1, left to its own devices, tends to reproduce that same flatness by summarizing metrics in the order given rather than by relevance, so explicitly instructing a re-rank by decision-relevance in Phase 1 is what forces the output to read as a narrative instead of a restated table. The instruction to say 'cause unconfirmed' rather than inventing an explanation directly targets a known failure mode of language models asked to explain metric movement: without a hard constraint, the model will construct a plausible-sounding causal story connecting the external factor to the number even when no such link was actually stated, because a confident causal narrative is a more satisfying completion than an honest gap — forcing the model to distinguish stated fact from inference keeps a fabricated cause from being presented with the same authority as a real one. Matching the register to the named audience matters mechanically because a model with no audience constraint defaults to a dashboard-analyst register full of internal shorthand, which reads as noise to a CFO or VP who doesn't touch the underlying data day to day. Ending with a stated confidence level on the recommendation, rather than a hedge-everything summary, is the difference between a review that produces a decision and one that produces another meeting to decide whether to decide — an explicit confidence level gives leadership something to either act on or explicitly challenge, instead of a non-committal 'it depends' that changes nothing.
What you get back
Headline: Churn worsened for the second straight month, and the timing lines up closely enough with the pricing increase that it's the leading suspect, though not confirmed. Churn: 4.2% vs 3% target, worse than last month's 3.1% — plausibly tied to the pricing change, cause unconfirmed. Recommendation (moderate confidence): hold off on a full rollback, but exempt the renewal cohort most exposed to the increase before next month's cycle.
Verified against
ChatGPT GPT-5.1 · 2026-08-11
Changelog
- 2026-08-11 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — custom software, built and maintained for you — that's Scult's day job.
EXPLORE CUSTOM SOFTWARE
