# Prompt 1 — Build an Executive Summary

> Adapted from Chapter 8 (Sara worked example, EMEA Q3 miss). The CSAR-loop entry point for prose-heavy summaries built from a semantic model.
>
> **Verified against**: Power BI Copilot, June 2026 surface

## Crystallize (the analyst's first move)

> "Before you build me anything, what would you need to know to write a defensible one-page summary of why EMEA missed plan in Q3?"

The Crystallize move forces Copilot to surface its missing-information list *before* committing to an output. The cost of asking is one turn; the cost of skipping is rework.

## Scope (Copilot's restatement)

Copilot will return a list of clarifying questions — in the Sara example, six of them: which definition of *plan*, which sub-regions to break out, what time grain, whether to include forward-looking forecast, what the audience-decision pairing is, and what off-the-table topics to avoid. Answer each in a sentence; let Copilot restate the agreed scope back.

The scope-restatement is the contract. If Copilot's restatement is wrong, fix it in the conversation — do not let the wrong scope get baked into the visual.

## Assemble

> "Build the one-page summary now."

Expect one page with two visuals (variance + trend) and a single takeaway sentence.

## Refine (corrections named by Part I principle)

- **Ch 3 (Big Idea)**: If the takeaway sentence reads as a temporary-cause framing when the data shows structural decline, rewrite the takeaway with the structural framing. The Big Idea must be defensible against the data, not optimistic against the audience.
- **Ch 6 (color accessibility)**: If the variance bar uses brand-palette red and green, run the color-blindness simulator (see [Chapter 6 companion](/ch/06/)) before ship.

## Ship decision

Ship when the headline number reconciles to the prior-period close to within rounding. If reconciliation fails, the work goes back to the data, not to the prompt.

---

**When to use this prompt**: Audience expects narrative-shaped output (executive summary, briefing memo, post-meeting recap) rather than a dashboard. The Crystallize-first move is what distinguishes the discipline from prompt roulette — see the [companion to Chapter 8](/ch/08/) for the underlying CSAR-loop methodology.

**Dataset**: [sara-emea-csar](/datasets/sara-emea-csar/) — download CSV + JSON to reproduce the Sara CSAR-loop pass end-to-end.