Market Audit

Market Audit — 6-Dimension Site Audit

Your job is to produce a rigorous marketing audit. You coordinate 5 specialist subagents running in parallel, then synthesize their outputs into one decision-grade report.

Phase 1 — Discovery (before spawning agents)

1. Fetch the page once with python3 scripts/analyze_page.py <url> and capture the JSON. Pass this data to every subagent so none of them re-fetches unnecessarily.
2. Classify the business type (SaaS / E-commerce / Agency / Local / Creator / Marketplace). This adjusts which findings matter most.
3. Extract the brand name from the title or logo alt text. Use it consistently in the report.
4. Check for prior artifacts (BRAND-VOICE.md, COMPETITOR-ANALYSIS.md, etc.) and feed relevant excerpts to the matching subagent.

Phase 2 — Parallel Analysis

Spawn these 5 subagents in the same message (one tool block with 5 Agent calls). Each gets: the URL, the analyze_page JSON, the detected business type, and the brand name.

Each subagent returns a 0–10 score per sub-dimension, specific findings with quoted evidence, and concrete fixes.

If a subagent fails or returns thin data, fall back to running the checks yourself from the analyze_page JSON — never leave a dimension blank.

Phase 3 — Synthesis

1. Normalize scores to 0–100 and apply the weights from the catalog.
2. Compute the composite and map to a letter grade.
3. Deduplicate findings — if Content and Conversion both flag the weak CTA, mention it once under the higher-impact dimension.
4. Rank every finding by revenue impact: impact = monthly_traffic × expected_CR_lift × ARPU. If traffic and ARPU aren't known, estimate conservatively and flag the assumption.
5. Generate the action plan in three tiers:
   • This week (quick wins) — under 1 day of work each
   • This month (high-leverage) — 1–3 weeks of work
   • This quarter (strategic) — 1–3 months of work

Output: MARKETING-AUDIT.md

# Marketing Audit — [Brand Name]
**URL:** [url]
**Business type:** [SaaS/Ecom/etc]
**Date:** [YYYY-MM-DD]

## Overall Score: [X]/100 — Grade [A-F]

[2–3 sentence executive summary: current health, top opportunity, estimated annual revenue impact of the recommendations.]

## Category Scores
| Category | Score | Weight | Status |
|---|---|---|---|
| Content & Messaging | X/100 | 25% | [Strong/OK/Weak] |
| Conversion Optimization | X/100 | 20% | ... |
| SEO & Discoverability | X/100 | 20% | ... |
| Competitive Positioning | X/100 | 15% | ... |
| Brand & Trust | X/100 | 10% | ... |
| Growth & Strategy | X/100 | 10% | ... |

## Critical Findings
[5–10 findings ordered by severity. Each has: severity label, one-line finding, evidence (quoted), and the fix.]

### 1. [Critical] Homepage headline is generic
**Evidence:** H1 reads "Welcome to Acme — The Platform for Modern Teams"
**Impact:** Visitors can't tell what you do within 5 seconds. Estimated 20–35% of bounces attributable.
**Fix:** Rewrite to specify the outcome and audience: "Cut reporting time 75% — automated analytics for B2B growth teams"

## Wins (keep doing)
[3–5 things the site does well, with specific examples.]

## Action Plan

### This Week — Quick Wins
1. [Specific action + expected impact]
...

### This Month — High-Leverage
1. ...

### This Quarter — Strategic
1. ...

## Methodology
Scored across 6 weighted dimensions (see `/market` skill for full rubric).
Automated data from `analyze_page.py`; qualitative analysis from 5 specialist subagents.

Quality Bar

• Every finding must quote evidence from the actual page. If you can't quote it, don't include it.
• Every fix must be specific enough that a junior marketer could execute it without follow-up questions.
• Never report a score without the underlying finding — scores without evidence are theater.
• If the overall score is below 40, lead with that reality. Clients deserve honesty.
• If the site is genuinely strong, say so. An 85 is not a failure of the audit — it's an honest result.