Research & analysis

Spreadsheet → Insight

The problem

You have the numbers and you can describe them, but you cannot say what they mean or what should change because of them.

Who it's for: Analysts, Managers, Founders, Marketers
What it produces: The three findings that matter, what caused them, and the recommended action with its risk.
What you need: A dataset or report export; The decision it should inform; What you expected to see
Tools: ChatGPT · Claude · Gemini
Level: Advanced
Format: 6 steps · Advanced

The workflow — 6 steps

Step 1
State your prediction first
Write down what you expect before you look. The gap between expectation and reality is where insight lives; without it you will just confirm yourself.
```
Before analysing, I predict: {your predictions}. Record these. Then, from the data below, tell me only which predictions were wrong and by how much — nothing else yet.
```

Step 2

Separate description from explanation

Get the descriptive facts locked down before any causal story. Models will otherwise invent a plausible cause immediately.

List the observable facts in this data with no interpretation: magnitudes, directions, outliers, gaps. Then, separately, list candidate explanations for each — clearly labelled as hypotheses, with what evidence would confirm or kill each.

4 more steps in the full workflow

The remaining steps include the exact prompts, the Notion template and the worked example. They're part of the AI Workflow System — a one-time purchase, no subscription, 30-day money-back guarantee.

Unlock the full workflow Try a free workflow first

What changes when you run it

Before

A dashboard export and a summary that says traffic is up 12% and conversion is down 3%.

After

Two predictions falsified, one supported causal explanation, a recommendation with its downside and the metric that would prove it wrong.

Free prompts used in this workflow

Prompts are the ingredients; this workflow is the recipe. Each of these is free to copy.

Browse the full free prompt library →