Imported from innovaigs/Vibe-Managing (skills/marketing/website-conversion-audit/SKILL.md). Install upstream with npx skills add innovaigs/Vibe-Managing --skill website-conversion-audit. Copyright stays with the author.

Skill: Website Conversion Audit (4 C's)

Purpose

Diagnoses why a website (usually the homepage or a key landing page) fails to turn visitors into customers, by scoring it against the four principles of conversion-centered design — Clarity, Attention, Congruence, Credibility — and returning a red/yellow/green scorecard plus a prioritized, specific fix list. It stops the founder from paying to send traffic to a page that leaks it.

When to Use

• The founder asks "why don't visitors convert?", "why is my bounce rate high?", "can you review my website / landing page?", "is my homepage any good?"
• Before spending on any channel that drives traffic to the site (paid ads, SEO) — the destination must be conversion-ready first.
• After a redesign or new landing page launch, to validate it before scaling traffic.
• When channel-selection names an owned page as the conversion destination and flags that it needs to be built or checked.

When NOT to Use

• There is no defined conversion goal for the page → first define primary + secondary conversions (this skill will ask; it cannot score congruence without a goal).
• The problem is "not enough traffic," not "traffic doesn't convert" → use channel-selection / keyword-and-search-map to drive traffic; this skill assumes traffic exists or is coming.
• The founder wants keyword/SEO coverage analysis → use keyword-and-search-map.
• The founder wants the value proposition itself written → use customer-value-proposition-builder; this skill only checks whether the USP is clearly expressed on the page.
• Deep technical performance/accessibility engineering (page speed internals, WCAG remediation code) → note the finding and hand to the appropriate specialist/tool; this skill flags them, it doesn't fix code.

Required Context

• customers.personas — who the page is for, their value_drivers, and what they came to accomplish.
• offerings.cvp_id / customer-value-proposition-builder output — the intended USP, to check the page expresses it.
• strategy / channel-selection output — the page's role and the traffic it will receive.
• goals — the business objective the conversion serves.
• metrics / integrations — current conversion rate, bounce, traffic sources if web analytics is connected (turns a qualitative audit into a measured one).
• Site type must be known or inferred: retail/eCommerce, service/lead-gen, media, or social/blog — it sets the expected primary/secondary conversions.

Inputs

input:
  page_url: str                    # REQUIRED. The page to audit.
  page_content: str | rendered     # REQUIRED. Text + structure of the page (or a capture).
  site_type:                       # REQUIRED (or inferred).
    type: enum(retail_ecommerce, service_leadgen, media, social_blog)
  target_persona_id: str           # REQUIRED. Who this page serves.
  primary_conversion:              # REQUIRED. The one action the page must drive.
    type: str                      # e.g. "purchase", "book a call", "request a quote"
  secondary_conversions: [str]     # Optional. e.g. create account, email signup, download
  intended_usp: str                # Optional. The value proposition it should express.
  current_metrics:                 # Optional, if analytics connected.
    type: {conversion_rate: number, bounce_rate: number, avg_time_on_page: number,
           top_traffic_source: str}
  device_scope:                    # Optional.
    type: enum(desktop, mobile, both)

Missing Information Protocol

1. No defined conversion goal → ask the founder ONE batched question: "What is the single most important action a visitor should take on this page, and what secondary actions would still be a win?" Do not audit without it — Congruence and Clarity are scored against the goal.
2. Site type unknown → infer from content and confirm; use it to load the expected conversion set (Retail → primary: Purchase; secondary: create account, email signup, premium upgrade. Service → primary: Contract/booking; secondary: request info, get a quote, download content).
3. No analytics connected → run a qualitative audit and clearly label scores as heuristic; add "connect web analytics to confirm" as a fix, and recommend re-audit once real conversion data exists.
4. Intended USP missing → still score Clarity/Credibility, but flag "USP not verifiable" and hand off to customer-value-proposition-builder if the page has no discernible differentiator.
5. Never assume the page converts because it "looks nice," never assume mobile is fine because desktop is, and never invent metric values — if analytics is absent, say so.

Diagnostic Questions

• What type of site/page is this, and what should the visitor accomplish here?
• What is the primary conversion goal, and what are the secondary conversions?
• On a 5-second scan, is it clear what the page is, who it's for, and what to do next? (Clarity)
• Is there ONE obvious primary action, or is attention split across many competing elements? (Attention)
• Does every headline, image, and word push toward the conversion, or do some pull away? (Congruence)
• Are there ample, specific reasons to believe you'll deliver — proof, social proof, guarantees? (Credibility)
• Is the unique selling proposition clear and above the fold?
• Is there information overload? Is the info ordered by importance? Do clickable elements stand out?
• Does the experience hold up on mobile as well as desktop?

Analysis Framework

Score the page on the 4 C's of Conversion-Centered Design, each rated Red / Yellow / Green, plus two supporting 1–5 ratings.

1. Clarity — Can a first-time visitor tell, within a quick scan, what this is, who it's for, and what to do? Essential information present and ordered by importance; no jargon wall.
2. Attention — Is focus directed to the one key action? Count distractors competing with the primary CTA; fewer is better. The primary CTA should be visually dominant and unmistakable.
3. Congruence — Do all elements (headline, subhead, imagery, copy, CTA label) reinforce the single conversion? Flag anything that dilutes or contradicts the goal (off-message links, competing offers, generic stock imagery).
4. Credibility — Are there ample, specific reasons to believe? Social proof (reviews, ratings, logos, testimonials), authority signals (credentials, awards, guarantees), and evidence the business delivers.

Supporting ratings (1–5):

• Information Overload (1 = overwhelming, 5 = clean) — is the page trying to do too much on one screen?
• USP Clarity (1 = absent, 5 = crystal clear and prominent) — is the differentiator obvious and above the fold?

Design rules applied to generate fixes (SOURCE): don't do too much on one page; make the USP clear; order information by importance; split dense content into multiple pages; make clickable elements stand out; keep text short; prefer one dominant action per page.

Calculations

Mostly qualitative scoring; supported by these where analytics exists.

• Website Conversion Rate = conversions ÷ site visitors. [CLAUDE-DERIVED — the source teaches the 4 C's qualitatively; the numeric conversion-rate benchmark is Claude-supplied.] Benchmark against site type rather than a universal number.
• Bounce rate and avg time on page — read from analytics if present; used to corroborate a Red Clarity/Attention score (high bounce + low time ⇒ confirms leak).
• C-score roll-up (CLAUDE-DERIVED heuristic) — map Green=2, Yellow=1, Red=0 across the 4 C's for an at-a-glance 0–8 readiness score; used only to prioritize, not published as an official metric.
• Expected lift framing (CLAUDE-DERIVED, directional) — when recommending a fix, estimate impact qualitatively (high/medium/low) rather than promising a specific % lift, unless A/B data exists.

Decision Rules

• IF any of the 4 C's scores Red THEN record a specific fix to bring it to Green and mark the page "not traffic-ready" until fixed — do not scale paid traffic to it. [SOURCE]
• IF any C scores Yellow THEN record a fix but allow traffic to proceed with the fix scheduled. [SOURCE]
• IF all 4 C's are Green THEN the page is traffic-ready; recommend measurement + iterative A/B testing rather than a rebuild. [SOURCE]
• IF Attention shows more than one co-equal primary CTA THEN flag competing calls-to-action; recommend demoting all but one primary action. [SOURCE]
• IF Information Overload ≤ 2 THEN recommend splitting content across multiple pages and cutting text. [SOURCE design rules]
• IF USP Clarity ≤ 2 THEN flag missing/weak differentiator and hand off to customer-value-proposition-builder. [SOURCE]
• IF Congruence is Red because off-message links pull attention THEN recommend removing or relocating them below the fold / to secondary pages. [SOURCE]
• IF Credibility is Red THEN recommend adding specific social proof (reviews/ratings/testimonials/logos) and any authority signals available. [SOURCE, ties to influence levers]
• IF site_type == retail_ecommerce THEN expect Purchase as primary; secondary = create account, email signup, premium upgrade. IF site_type == service_leadgen THEN expect Contract/booking as primary; secondary = request info, quote, content download. [SOURCE]
• IF no analytics connected THEN label all scores heuristic and add "connect analytics + set up conversion tracking" as a prerequisite fix. [SYNTH]
• IF device_scope != mobile and traffic is largely mobile THEN require a mobile pass before sign-off. [SYNTH]
• IF a fix requires publishing/replacing live public content THEN hold the actual publish for founder approval; the audit and fix drafts do not. [POLICY]

Procedure

1. Confirm goal & type — lock primary + secondary conversions and site type; load the expected conversion set.
2. Load persona & USP — know who the page serves and the differentiator it should express.
3. Pull metrics — if analytics connected, read conversion rate, bounce, time-on-page, top source to corroborate findings.
4. 5-second scan test — assess Clarity first as a first-time visitor would.
5. Score each C Red/Yellow/Green with a one-line evidence note per score.
6. Rate Information Overload and USP Clarity (1–5).
7. Mobile pass if in scope.
8. Generate fixes — for every Yellow/Red, a specific, actionable fix mapped to the design rules; tag each fix with impact (H/M/L), effort, and whether it needs approval to publish.
9. Prioritize — Reds first (blocking), then Yellows by impact/effort; produce a traffic-readiness verdict.
10. Write back — save the scorecard + fix list to strategy, log a decision record, and (if analytics present) snapshot the baseline conversion metric to metrics. Hand off to customer-value-proposition-builder (weak USP), keyword-and-search-map (search-landing pages), and marketing-metrics-tracker (to track post-fix conversion).

Output

output:
  page_url: str
  site_type: enum(retail_ecommerce, service_leadgen, media, social_blog)
  primary_conversion: str
  secondary_conversions: [str]
  scorecard:
    clarity:     {rating: enum(red, yellow, green), evidence: str}
    attention:   {rating: enum(red, yellow, green), evidence: str}
    congruence:  {rating: enum(red, yellow, green), evidence: str}
    credibility: {rating: enum(red, yellow, green), evidence: str}
  information_overload: number     # 1-5
  usp_clarity: number              # 1-5
  c_score: number                  # 0-8 roll-up, heuristic
  traffic_ready: bool
  measured: bool                   # true if backed by analytics
  baseline_metrics:                # present only if measured
    type: {conversion_rate: number, bounce_rate: number, avg_time_on_page: number}
  fix_list:
    - fix: str
      addresses: enum(clarity, attention, congruence, credibility, overload, usp, mobile, tracking)
      severity: enum(blocking, high, medium, low)
      impact: enum(high, medium, low)
      effort: enum(low, medium, high)
      needs_publish_approval: bool
  verdict: str                     # plain-language summary + go/no-go on traffic
  handoffs: [str]
  confidence: enum(low, medium, high)

Recommendations

Fixes are prioritized blocking-first (any Red that makes the page not traffic-ready), then by impact ÷ effort. Reversible edits that don't touch live content (copy drafts, layout mockups, adding staged social proof) are recommended for immediate preparation; anything that publishes to the live public site is presented as an approval-gated action. The single highest-leverage fix is called out explicitly, and the founder is told the one thing to fix before spending on traffic.

Execution Opportunities

• Draft the scorecard and fix list; write to strategy — reversible, L1.
• Draft revised copy / CTA labels / page structure as proposals (not published) — reversible, L1.
• Create internal tasks for each fix with owner and effort — reversible, L2 candidate.
• Set up / request conversion tracking in connected analytics — configuration change, approval-gated.
• Snapshot the baseline conversion metric to metrics for before/after comparison — reversible, L2.
• Schedule a re-audit / A/B test after fixes ship — reversible, L2.

Human Approval Requirements

• Publishing or replacing any live public page content → ALWAYS founder approval (external-facing, brand-critical).
• Changing analytics/tracking configuration or standing site settings → founder approval.
• Adding testimonials/reviews/logos → founder approval AND verify permission/authenticity (never fabricate social proof).
• The audit itself, drafts, mockups, and internal tasks require no approval (L1/analysis).

Escalation Conditions

• Legal/regulatory claims on the page (health, financial, guarantees, comparative claims) → Legal Liaison agent before publishing changes.
• Accessibility or serious technical performance defects beyond copy/layout → flag to an engineering/design specialist.
• No USP exists at all (not just unclear) → escalate to customer-value-proposition-builder + founder; a conversion audit can't fix an absent value proposition.
• Fixes require budget (redesign, developer time) → founder + CFO for cost.

KPIs

• Post-fix website conversion rate improvement vs. baseline (the core measure; CLAUDE-DERIVED benchmark by site type). [primary]
• Reduction in bounce rate / increase in time-on-page for the audited page.
• Number of Red items closed to Green; % of fix list shipped.
• Whether the page passed all 4 C's before paid traffic was scaled (leak prevented).

Monitoring

After fixes ship, watch conversion rate, bounce, and time-on-page for the page (via marketing-metrics-tracker); compare against the baseline snapshot. If a fix didn't move the metric, re-open the corresponding C for a deeper look or an A/B test. Watch mobile vs. desktop conversion separately.

Follow-Up

• Event-triggered: any redesign, new landing page, new campaign pointing traffic at the page, or a drop in conversion flagged by monitoring.
• Time-triggered: re-audit each time channel-selection designates the page as a paid-traffic destination, and on a periodic cadence for high-traffic pages.

Related Skills

• channel-selection — designates the owned page as a conversion destination and triggers this audit before spend.
• customer-value-proposition-builder — supplies/repairs the USP the page must express.
• keyword-and-search-map — for search-landing pages, ensures the page also matches search intent.
• marketing-funnel-planner — the page is the "tool" for the Interest/Action funnel stage.
• marketing-metrics-tracker — measures conversion before and after fixes.

Guardrails

• Never publish changes to a live public page without founder approval.
• Never fabricate or add unverifiable social proof, reviews, testimonials, or credentials — credibility fixes must use real, permitted evidence.
• Do not promise a specific conversion-rate lift; give directional impact unless A/B data supports a number.
• When analytics is absent, clearly mark scores as heuristic; do not present unmeasured judgments as measured facts.
• Do not overwrite the page's existing content in memory; store fixes as proposals.
• Flag regulated or legal claims for review before they ship.

Example

Founder input: "Here's my homepage. I'm about to run ads to it but almost no one books a call. Can you check it?" (Service/lead-gen; persona = small-clinic office manager; primary conversion = "book a discovery call"; analytics connected: conversion 0.6%, bounce 74%.)

Skill reasoning:

• 5-second scan → the headline is a generic slogan ("Excellence in Healthcare Solutions"); unclear what's offered or for whom → Clarity: Red.
• Three co-equal buttons ("Book a Call", "Read our Blog", "See Pricing") plus a newsletter popup → attention split → Attention: Red (>1 primary CTA + distractor popup).
• Hero image is generic stock; body copy talks about the company's history, not the visitor's problem → some elements pull away from booking → Congruence: Yellow.
• No reviews, no client logos, no guarantee; one vague "trusted by many" line → Credibility: Red.
• Information Overload = 2 (dense history section); USP Clarity = 1 (no differentiator visible). Bounce 74% + conversion 0.6% corroborate the Clarity/Attention Reds.
• Verdict: NOT traffic-ready — 3 Reds. c_score = 1/8.

Output (abridged):

• Fixes (prioritized): (1) blocking, high impact/low effort — rewrite headline to name the offer + audience + outcome ("Fewer no-shows for small clinics — book a 15-min call"); addresses Clarity/USP. (2) blocking, high/low — make "Book a Call" the single dominant CTA; remove the newsletter popup and demote blog/pricing to the nav; addresses Attention. (3) blocking, high/medium — add 3 real client testimonials + logos + a satisfaction guarantee; addresses Credibility (needs_publish_approval + verify authenticity). (4) high, low — replace stock hero with a product/outcome visual; cut the history section (Congruence, Overload). (5) medium — confirm mobile layout.
• Handoffs: customer-value-proposition-builder (USP=1), marketing-metrics-tracker (track post-fix conversion).

Executed vs. approval: Drafted the scorecard, rewritten headline/CTA copy, and internal fix tasks with no approval. Snapshotted the 0.6% baseline. The live publish of the new copy and the testimonials was held for founder approval (external content + must verify testimonial authenticity), and the founder was advised not to launch ads until the three Reds are Green.

Provenance

SOURCE. Derived from the Marketing & Customer domain knowledge (Conversion-Centered Design "4 C's" — Clarity/Attention/Congruence/Credibility; Red/Yellow/Green scoring; Information-Overload and USP-clarity ratings; site-type→conversion-goal reference; homepage design rules; the "fix red/yellow before spending on traffic" rule). The numeric website-conversion-rate benchmark, the 0–8 c-score roll-up, and directional-lift framing are CLAUDE-DERIVED and flagged inline. See internal/PROVENANCE_MAP.md.