Verified against Claude · 2026-07-29
Turn raw campaign metrics into a plain-English performance takeaway
Feed in numbers you've already calculated — spend, revenue, ROAS, ROI — and get a written takeaway a founder or client can read in two minutes. This does not calculate anything: run your numbers through the Marketing ROI Calculator first, then bring the output here.
The prompt
Ready to copy — highlighted parts are example details you can swap.
I've already calculated these numbers for the campaign "August Meta Prospecting — New Customers" over August 1-31, 2026 — do not recalculate or second-guess the math, treat every figure below as correct: Spend ₹2,80,000. Revenue attributed ₹9,10,000. ROAS 3.25x. Gross margin 30%. ROI (from Marketing ROI Calculator) -2%. Conversions 145. CPA ₹1,931. Previous month ROI was +18%. This will be read by: Founder who only has 2 minutes and cares about profit, not ROAS. Write a performance takeaway, not a re-explanation of the numbers: 1. One-sentence verdict — did this campaign make or lose money, in plain terms 2. The single most important reason why (point at the specific number that explains it — e.g. a gap between ROAS and ROI, a CPA trend, a conversion-rate shift) 3. What changed versus the previous period, if I gave you that comparison, and whether that change is the headline or a side note 4. One thing worth double-checking before acting on this (e.g. attribution window, whether margin is accurate, seasonality) — not a hedge on every line, just the one caveat that actually matters here 5. A recommended next action, phrased as a decision (continue / pause / adjust budget / adjust targeting), not a vague "monitor closely" Do not introduce any number I didn't give you. If you need a number I haven't provided to make the verdict solid, ask for it instead of estimating it.
Customize the highlighted detailsoptional — the prompt above already works
Why this works
The failure mode this avoids is an LLM quietly inventing or 'helpfully' recalculating a benchmark figure it wasn't given — a real financial verdict should come from an actual calculation (the Marketing ROI Calculator's documented ROI formula, not a model doing arithmetic from a paragraph), and this prompt is deliberately scoped to interpretation and narrative only. Locking the model to the numbers you supply, and instructing it to ask rather than estimate, keeps the output useful for what an LLM is actually good at — turning a spreadsheet of numbers into the one sentence a founder needs — without letting it drift into inventing the underlying math.
What you get back
Verdict: This campaign lost money in August despite a 3.25x ROAS. The gap is margin: at 30% gross margin, break-even ROAS is about 3.3x, so 3.25x lands just under water — ROI of -2% confirms it. Versus July's +18% ROI, this is the headline, not a side note. Before acting: confirm the attribution window didn't shift between months, since that alone could explain part of the swing. Recommended action: pause and re-test creative before adding budget — don't scale a campaign that's currently losing money even at a healthy-looking ROAS.
Verified against
Claude Claude Opus 4.5 · 2026-07-29
ChatGPT GPT-5.1 · 2026-07-30
Changelog
- 2026-07-30 — Initial version, explicitly scoped to narrative-only after testing showed models will "helpfully" recompute ROI from partial inputs unless told not to.
Need this built into your business?
If a prompt isn't enough — Google Ads management, built and maintained for you — that's Scult's day job.
EXPLORE GOOGLE ADS MANAGEMENT
