---
name: marketing_performance_brief
description: "Write the daily or weekly marketing performance brief: consolidate search, paid ads, social, and CRM/email metrics into a one-page read on what changed, why it changed, and what to do next. Uses real metric definitions and diagnostic decision rules, works with any stack (Google Ads, Meta Ads, Search Console, GA4, Mailchimp/Klaviyo, any CRM). Triggers on: marketing brief, weekly marketing report, daily performance update, campaign performance summary, marketing metrics review."
---

# Marketing Performance Brief

## Purpose

Turn a pile of per-channel numbers into one page a busy person can actually read. The brief answers three questions, in this order: what changed, why it changed, what to do next. It covers four inputs: paid search, paid social, organic search, and email/CRM. It is tool-agnostic: the same brief works whether the numbers come from Google Ads, Meta Ads Manager, Search Console, GA4, Mailchimp, Klaviyo, or a CRM export.

## Workflow

1. **Gather the canonical metric set per channel.** Collect one period of data (day or week) plus the prior equivalent period as baseline. Use the channel metric tables in `references/metrics.md` so every metric is defined the same way every time. Refuse to brief on a channel where you cannot get the volume metrics (spend, sends, impressions): rates without volumes are noise.

2. **Compute movers, not everything.** For each metric, compute the period-over-period change in percent. A metric becomes a "mover" only if it moved more than the noise floor for that channel's spend volume: high-spend accounts react to 15-20% swings; low-spend or high-variance accounts treat anything under ~30% as noise. Flag each mover as good, bad, or neutral before writing a word.

3. **Diagnose with the decision rules.** Run each mover through `references/decision-rules.md`. The rules are directional starting points, not laws: they name the most likely cause of a pattern (creative fatigue vs. audience decay vs. tracking breakage) and the specific next action. If a pattern matches no rule, say so and propose the cheapest test that would identify it.

4. **Cross-channel check.** Compute the blended view: total spend, total attributed revenue, MER (marketing efficiency ratio = total revenue / total ad spend), and blended CAC. A channel can look worse while the whole looks better (or the reverse). Never report a channel win that the blended number contradicts without explaining the gap.

5. **Draft the one-page brief.** Follow the Output Contract below and the worked template in `references/brief-template.md`. Keep the whole brief to one page (or one screen). If it spills over, cut a chart, never a decision.

## Output Contract

The brief has exactly five sections, in order:

1. **Headline.** One sentence: the period, the verdict (on track / needs attention / off track), and the single biggest driver. Example: "Week of Oct 5: on track. Paid search drove the gain, up 9% in conversions on flat spend."
2. **Scoreboard.** One compact table: channel rows, columns for spend/sends, key rate metric, outcome metric (CPA/CAC or ROAS), and period-over-period change with the baseline named. Every rate metric appears next to its volume metric.
3. **What changed.** Bullets, one per mover, each with the metric, the numbers, and the percent change. No interpretation here, just facts.
4. **Why.** For each mover, the diagnosis from the decision rules: the likely cause and the evidence for it. When the evidence is thin, say so: "No clear cause; needs a test."
5. **What to do next.** Numbered actions, each with an owner slot and a check date. Max five actions. Anything that would not get done before the next brief is not an action, it is a wish.

Closing line: what to watch before the next brief (the one metric that would flip the verdict).

## Operating Rules

1. **Every number earns its place.** No invented numbers, dates, or quotes, ever. Benchmarks come from the cited sources in `references/metrics.md` or are explicitly labeled as account targets set by the reader. Unverifiable claims are cut or framed as opinion.
2. **Rates never travel alone.** CTR without impressions, open rate without delivered count, ROAS without spend: each is a half-truth. Pair every rate with its volume.
3. **Judge CTR at a position, never in a vacuum.** A 2% organic CTR is terrible at position 2 and fine at position 9. Always name the position or placement beside the rate.
4. **Attribution gets one honest sentence.** If the numbers mix 7-day-click, last-click, and platform-attributed revenue, say so up front and do not add them into a fake total. The blended view uses one consistent attribution source or is skipped.
5. **Decision rules are triage, not diagnosis.** A frequency over 3 with falling CTR suggests creative fatigue; it does not prove it. The brief proposes the check that would confirm, it does not declare certainty it lacks.
6. **Boring and true beats clever.** Plain sentences, real rhythm, no hype. No em dashes. If a sentence sounds like a press release, rewrite it before it ships.