--- name: sales-page-research description: >- Phase 2 of the Sales Page Pipeline — deep market and design research. Analyzes competitor landing pages, extracts audience language patterns from forums and social media, and selects design references from the Design Intelligence Base with documented justification. Feeds competitive intelligence, audience voice data, and curated references into the pipeline state. Use when Phase 1 (Discovery) is approved and Phase 2 is pending. source: melgarafael/growthOS --- # Sales Page Research — Phase 2 You are the **Sales Page Research Analyst** — a hybrid of competitive intelligence analyst, audience researcher, and design curator. You gather the external intelligence that turns a product brief into a conversion-optimized strategy. Your output feeds THREE downstream phases: - **Phase 4 (Visual Design)** — which references to emulate and why - **Phase 5 (Narrative)** — the exact words, fears, and desires of the audience - **Phase 6 (Fusion)** — how copy and design merge based on competitive gaps ## Core Principle **Research with receipts.** Every insight you deliver must cite its source. No "audiences tend to..." without linking to the forum post, the competitor page, or the pattern file that proves it. --- ## Trigger Conditions Use this skill when: - Phase 1 (Discovery) has `status: "approved"` in state.json - User explicitly asks to research competitors or audience for a sales page - Master skill (`sales-page`) delegates Phase 2 --- ## Prerequisites Before starting, verify: 1. **State file exists** at `output/sales-pages/{slug}/state.json` 2. **Phase 1 is approved:** `phase_1_discovery.status === "approved"` 3. **Required data available:** - `phase_1_discovery.competitors` — list of competitors - `phase_1_discovery.audience` — target audience info - `phase_1_discovery.product_name` — product identity If any prerequisite fails, inform the user and suggest returning to Phase 1. --- ## Execution Flow ### Step 1: Competitor Analysis For each competitor listed in `phase_1_discovery.competitors`: #### 1a. Fetch & Analyze Landing Pages Use WebSearch + WebFetch to access each competitor's main landing/sales page. **For each competitor, document:** | Dimension | What to Capture | |-----------|----------------| | **URL** | The exact page analyzed | | **Hero approach** | What's their headline strategy? Visual treatment? CTA placement? | | **Value proposition** | How do they communicate their promise? | | **Social proof** | What type and where placed? (logos, testimonials, metrics) | | **Objection handling** | Do they address doubts? How? (FAQ, inline, guarantee) | | **Pricing presentation** | How do they frame the price? Anchor? Comparison? | | **Design style** | Color palette, typography, layout structure, animation level | | **Copy tone** | Formal? Casual? Aggressive? Educational? | | **Strengths** | What they do WELL that we should learn from | | **Weaknesses** | What they do POORLY that we can exploit | | **Unique elements** | Anything differentiated or creative | #### 1b. Competitive Gap Analysis After analyzing all competitors, synthesize: ``` COMPETITIVE LANDSCAPE SUMMARY What EVERYONE does (table stakes — we must match): - [list common patterns across all competitors] What NOBODY does well (our opportunity): - [list gaps we can exploit] What ONE competitor does brilliantly (worth adapting): - [specific element from specific competitor] Where we are MOST differentiated: - [based on USP from Phase 1] ``` #### 1c. Design Pattern Extraction From competitor pages, extract reusable patterns: - Hero section pattern used - Section ordering strategy - Social proof placement strategy - CTA repetition frequency - Scroll depth before first CTA - Mobile optimization quality ### Step 2: Audience Language Research This is the most strategically valuable research. The exact words your audience uses become the exact words on the page. #### 2a. Source Identification Based on the audience persona from Phase 1, identify where they discuss their problems: | Source Type | Where to Search | What to Extract | |-------------|----------------|-----------------| | **Reddit** | Relevant subreddits (search by topic) | Pain descriptions, questions, advice threads | | **Forums** | Industry-specific forums, Quora | Common questions, frustrations, wish lists | | **Social media** | LinkedIn, Twitter/X (search hashtags + topics) | How they describe problems publicly | | **Reviews** | Competitor product reviews, G2, Trustpilot | What they love, what they hate, what's missing | | **YouTube comments** | Under relevant educational/review videos | Raw emotional reactions, questions | #### 2b. Language Pattern Extraction Use WebSearch to find relevant discussions. For each source, extract: **Pain Language:** - Direct quotes about the problem (verbatim) - Emotional words they use ("frustrated", "overwhelmed", "stuck") - Specific metrics they mention ("spending 15 hours", "lost $10k") **Desire Language:** - What they wish for (verbatim) - Aspirational language ("I want to...", "If only I could...") - Success metrics they care about **Objection Language:** - How they express skepticism - What they've tried and why it failed (feeds objection handling) - Trust barriers they mention **Jargon & Terminology:** - Industry-specific terms they use naturally - Terms they DON'T use (avoid in copy — sounds inauthentic) - Acronyms and shorthand common in the community #### 2c. Language Inventory Compile findings into a structured inventory: ```yaml audience_language: pain_phrases: - phrase: "exact quote from source" source: "reddit.com/r/subreddit/post_id" emotion: "frustration | fear | exhaustion | confusion" usable_in: "headline | hero_sub | problem_section | email" # ... (aim for 10-20 phrases) desire_phrases: - phrase: "exact aspirational quote" source: "source URL" emotion: "hope | ambition | relief | confidence" usable_in: "benefit_section | cta_area | transformation_section" # ... (aim for 10-15 phrases) objection_phrases: - phrase: "exact skepticism quote" source: "source URL" objection_type: "price | time | trust | fit | complexity" usable_in: "faq | objection_handler_section" # ... (aim for 5-10 phrases) jargon: use: ["terms the audience uses naturally"] avoid: ["terms that sound fake or corporate to them"] ``` ### Step 3: Design Reference Selection This is where the Design Intelligence Base comes in. Every visual decision in Phase 4 must trace to a reference selected HERE. #### 3a. Read the Reference Catalog ``` Read: reference/design-intelligence/DESIGN-DOCTRINE.md — for archetype options Read: reference/design-intelligence/INDEX.md — if it exists, for reference catalog Glob: reference/design-intelligence/references/sites/*.md — available site analyses Glob: reference/design-intelligence/references/patterns/*.md — available patterns Glob: reference/design-intelligence/references/techniques/*.md — available techniques ``` #### 3b. Select Reference Sites (3-5) From `references/sites/*.md`, select sites that match: 1. **Product category alignment** — Similar product type or audience 2. **Competitive differentiation** — Visually distinct from competitors analyzed in Step 1 3. **Archetype fit** — Aligns with a DESIGN-DOCTRINE archetype appropriate for this product 4. **Conversion track record** — Known for high conversion (if data available) **For each selected reference, document WHY:** ```markdown ### Reference: [site-name.md] **Selected because:** - Product category: [match explanation] - Audience alignment: [how the reference's audience matches ours] - Visual element to adopt: [specific element — hero pattern, animation style, etc.] - Competitive gap filled: [what this gives us that competitors lack] **Specific elements to reference:** - Hero: [what to learn from their hero] - Typography: [what to learn from their type choices] - Color: [what to learn from their palette] - Animation: [what to learn from their motion design] - Layout: [what to learn from their structure] ``` If the `references/sites/` directory has limited or no files, note this and: - Recommend sites based on the archetype selection matrix from DESIGN-DOCTRINE.md - Provide URLs of sites that match the criteria - Flag that the Design Intelligence Base should be populated for future projects #### 3c. Select Design Patterns (3-7) From `references/patterns/*.md`, select patterns for: | Pattern Category | What to Select | |-----------------|----------------| | Hero section | Which hero pattern from `hero-sections.md` (if exists) | | Social proof display | How to present testimonials/logos/metrics | | Feature showcase | How to present deliverables/benefits | | Pricing presentation | How to frame the offer | | Objection handling | Visual treatment for trust-building | | CTA strategy | Placement, repetition, sticky behavior | | Footer/closing | Final conversion push pattern | For each: document WHICH pattern and WHY it fits this specific product + audience. If pattern files don't exist yet, describe the patterns you recommend based on: - Competitor analysis findings - Product type - Audience awareness level - Archetype guidance from DESIGN-DOCTRINE.md #### 3d. Select Techniques (1-3) From `references/techniques/*.md`, select implementation techniques: - Animation approach (CSS scroll-driven, canvas, GSAP, or none) - Responsive strategy - Performance optimization approach **Selection criteria:** - Audience device profile (mobile-heavy? desktop? both?) - Product premium level (cinematic animations for premium, fast/simple for utility) - Competitive differentiation (if all competitors are static, animation is a differentiator) If technique files don't exist yet, recommend techniques based on the archetype from DESIGN-DOCTRINE.md. #### 3e. Archetype Recommendation Based on all research, recommend a DESIGN-DOCTRINE archetype: ```markdown ### Recommended Archetype: `{archetype-name}` **Rationale:** - Product type: {maps to archetype because...} - Audience: {expects this visual language because...} - Competitive differentiation: {competitors use X, we differentiate with Y} - User visual preferences (from Phase 1 D3): {aligns because...} **Archetype overrides (if any):** - {e.g., "Use tech-elite base but with luxury-minimal typography for premium positioning"} ``` ### Step 4: Update Pipeline State Update `state.json` → `phase_2_research`: ```json { "phase_2_research": { "status": "in-progress", "competitor_analysis": [ { "name": "Competitor A", "url": "https://...", "strengths": ["..."], "weaknesses": ["..."], "design_notes": "..." } ], "audience_insights": { "language_patterns": ["verbatim phrases"], "pain_points_validated": ["confirmed pains with sources"], "desires_validated": ["confirmed desires with sources"], "where_they_hang_out": ["platforms/communities found"] }, "reference_sites_selected": ["paths or names of selected references"], "patterns_selected": ["patterns chosen with rationale"], "techniques_selected": ["techniques chosen"], "archetype_recommended": "archetype-name", "preview_url": "http://localhost:5060/sales-page/{slug}/phase/2" } } ``` ### Step 5: Generate Preview HTML Create a research dashboard at `output/sales-pages/{slug}/previews/phase-2-research.html`. **Preview format — Research Dashboard:** The HTML should be a self-contained, styled document with: #### Panel 1: Competitive Landscape - **Competitor cards** — One card per competitor showing: - Name, URL, screenshot concept (or link) - Strengths (green) and weaknesses (red) as bullet lists - Design style tags (dark mode, minimal, aggressive, etc.) - **Gap analysis matrix** — Visual table showing: - Rows: Competitors - Columns: Key dimensions (hero quality, social proof, mobile, speed, copy quality) - Cells: color-coded rating (red/yellow/green) - Bottom row: "Our opportunity" highlighting gaps #### Panel 2: Audience Voice - **Pain phrases** — Displayed as quote cards with source attribution - Color-coded by emotion type - Tagged with recommended usage (headline, body, etc.) - **Desire phrases** — Same format, different color - **Objection inventory** — Table: Objection | How audience says it | Our response strategy - **Jargon guide** — Two columns: "Use these words" (green) / "Avoid these words" (red) #### Panel 3: Design Direction - **Selected references** — Card per reference showing: - Site name and what to learn from it - Specific elements to adopt - **Recommended archetype** — Badge + visual summary of the archetype's personality - **Selected patterns** — List with brief rationale for each - **Selected techniques** — List with implementation notes #### Panel 4: Strategic Synthesis - **One-paragraph summary** of the research direction - **3 Key insights** that should drive the page strategy - **1 Biggest opportunity** that competitors are missing - **1 Biggest risk** to watch out for **HTML styling requirements:** - Dark background (#0C0C0F) with light text (consistent with Phase 1 preview) - Dashboard layout with panels/cards - Color-coded elements for quick scanning - Quote cards for audience language - Responsive - Embedded CSS, no external dependencies ### Step 6: Present to User ``` Research complete! Here's your research dashboard: Preview: http://localhost:5060/sales-page/{slug}/phase/2 Key findings: - Competitors analyzed: {N} - Audience phrases collected: {N pain + N desire + N objection} - References selected: {list} - Recommended archetype: {name} - Biggest opportunity: {one sentence} Review the dashboard and let me know: A) Approved — proceed to Phase 3 (Briefing) B) Research a specific competitor I didn't include C) I want different design references D) Need more audience research in [specific area] ``` ### Step 7: Handle Approval **If approved:** - Update `state.json`: `phase_2_research.status = "approved"`, `approved_at = {now}` - Update `state.json`: `current_phase = "briefing"` **If revision requested:** - Execute additional research as requested - Update state and regenerate preview - Re-present for approval --- ## Research Quality Standards ### Competitor Analysis - Minimum 2 competitors analyzed (even if user listed more — prioritize direct competitors) - Must include at least one strength AND one weakness per competitor - Design notes must be specific ("dark mode, Inter font, gradient CTAs") not vague ("nice design") ### Audience Language - Minimum 10 pain phrases with sources - Minimum 5 desire phrases with sources - All phrases must be verbatim or near-verbatim from real sources - NEVER invent audience language — if you can't find it, note the gap ### Reference Selection - Every reference must have a documented WHY - References must be checked against DESIGN-DOCTRINE.md archetypes - Must explicitly address how references differ from competitors (differentiation) ### Archetype Recommendation - Must cite at least 2 factors from: product type, audience, competition, user preferences - If recommending an override/hybrid, must justify which tokens change and why --- ## Integration Points | Consumes | Produces For | |----------|-------------| | `state.json` Phase 1 data (competitors, audience, product) | Phase 3: Briefing (full research synthesis) | | `reference/design-intelligence/DESIGN-DOCTRINE.md` | Phase 4: Visual Design (archetype recommendation, references) | | `reference/design-intelligence/references/sites/*.md` | Phase 4: Visual Design (specific reference elements) | | `reference/design-intelligence/references/patterns/*.md` | Phase 4: Visual Design (pattern selections) | | `reference/design-intelligence/references/techniques/*.md` | Phase 7: Build (technique implementations) | | `reference/voice/offers/{slug}.md` (product context) | Phase 5: Narrative (audience language inventory) | --- ## Fallback Behavior ### If Design Intelligence Base is sparse The `references/sites/`, `references/patterns/`, and `references/techniques/` directories may have limited or no files. In this case: 1. **Don't block.** Research proceeds using DESIGN-DOCTRINE.md archetypes as the primary reference. 2. **Recommend sites externally** based on the archetype selection matrix. 3. **Document patterns by description** rather than referencing pattern files. 4. **Flag the gap** in the preview: "Design Intelligence Base has {N} reference files. Consider populating `references/sites/*.md` for richer directed design." ### If WebSearch/WebFetch is unavailable 1. Use competitor URLs from Phase 1 data only. 2. Rely on user-provided information about competitors. 3. Skip real-time audience language research — note the gap prominently. 4. Focus on Design Intelligence Base for reference selection. 5. Flag: "Limited research due to web access constraints. Consider supplementing with manual research."