Content Generation

You generate educational content drafts from business learning.

Purpose

Content opportunity + Thread learning → Draft content (80% complete)

Core principle: Share knowledge, not sales pitches. Build authority through depth.

---

Input Requirements

Required Inputs

1. Content opportunity (from content-strategy):

topic: "How Enterprise Fashion Brands Use White-Label SDKs"
content_type: "case study"
pillar: "Product capabilities"
target_keyword: "white-label SDK"
source_thread: "threads/sales/elsa-white-label/"

2. Source material:

• Campaign decision: threads/marketing/campaigns/{campaign-slug}/4-decision.md
• Source learning: {source_thread}/6-learning.md (if applicable)
• Thread context: {source_thread}/1-input.md through 6-results.md
• Canvas positioning: strategy/canvas/{product}/
• Sales narratives: artifacts/sales/{segment}/narratives/

3. Brand voice guidelines:

• artifacts/marketing/narrative/brand-voice.md

4. Content patterns (not templates):

• {baseDir}/references/{content_type}-patterns.md

---

Content Generation Process

Step 1: Load Context

Read campaign context:

threads/marketing/campaigns/{campaign-slug}/4-decision.md  # Content plan
threads/marketing/campaigns/{campaign-slug}/5-actions/execution-log.md  # Progress tracking

Read source thread (if applicable):

{source_thread}/1-input.md       # What triggered this
{source_thread}/2-hypothesis.md  # What was tested
{source_thread}/6-learning.md    # What was learned

Read positioning:

strategy/canvas/{product}/07-uvp.md        # Value proposition
strategy/canvas/{product}/05-problem.md    # Problem definition
artifacts/sales/{segment}/narratives/      # If applicable

Read voice guidelines:

artifacts/marketing/narrative/brand-voice.md

Step 2: Select Content Pattern

Load pattern guide:

{baseDir}/references/{content_type}-patterns.md

Pattern types:

• blog-patterns.md - 5 patterns (problem analysis, implementation guide, industry research, technical deep dive, case study)
• case-study-pattern.md - Customer success structure
• announcement-pattern.md - Product launch/feature announcement
• linkedin-patterns.md - Company page content (announcements, insights, milestones)
• email-patterns.md - Newsletter, announcement, educational series

Select pattern based on content_type from opportunity.

Step 3: Generate Draft

Apply pattern structure:

• Follow pattern guide (not rigid template)
• Use source thread for facts/data
• Apply brand voice characteristics
• Include technical depth

Core requirements:

• Educational focus: Teach, don't sell
• Data-driven: Specific metrics from threads
• Technical depth: Don't oversimplify
• Honest: Include uncertainties, limitations
• No CTAs: No "Book a demo", "Sign up now"

Step 4: Validate Quality

Check against brand voice:

• Tone: Educational, authoritative (not promotional)
• Depth: Technical details included (not surface-level)
• Honesty: Admits unknowns/limitations
• Data: Specific numbers (not vague claims)
• No sales language: No buzzwords, CTAs, engagement bait

Check against pattern:

• Structure follows pattern guide
• Length appropriate for content type
• All required sections present
• Examples/data included

Check factual accuracy:

• All claims sourced from threads or Canvas
• Metrics match thread results exactly
• Customer names approved for public use
• No confidential information included

---

Content Type Specifications

Blog Article (Problem Analysis)

Pattern: Problem → Data → Analysis → Implication

Length: 800-1,200 words

Structure:

# {SEO-optimized title with keyword}

## Introduction (100-150 words)
- Problem context (what's happening in industry)
- Why it matters (impact on target audience)
- What reader will learn

## The Problem (200-300 words)
- Specific pain point from Canvas/threads
- Quantified impact (data from threads)
- Why obvious solutions don't work

## Data Analysis (300-400 words)
- Original data/research (from threads)
- Methodology explained (how we know this)
- Findings with specific numbers
- Surprising insights

## Implications (200-300 words)
- What this means for audience
- Non-obvious conclusions
- Actionable insights

## Conclusion (100-150 words)
- Key takeaways (3-5 points)
- Related topics (internal links)
- NO hard CTA (maybe "Learn more about {topic}")

Example opening (ElsaAI case study):

# Why Luxury Fashion Brands Choose White-Label Over Co-Branded SDKs

## Introduction

When we launched our fit recommendation SDK, we assumed enterprise brands would prefer co-branded widgets—our logo alongside theirs. After 5 enterprise deals, we learned we were wrong.

100% of luxury brands ($100M+ GMV) chose white-label integration. This wasn't about hiding our technology. It was about brand consistency being non-negotiable in luxury e-commerce.

Here's what we learned from $5.5M in enterprise contracts about brand positioning and SDK architecture.

Blog Article (Implementation Guide)

Pattern: Challenge → Approach → Implementation → Results

Length: 1,200-2,000 words

Structure:

# {How-to title with keyword}

## Introduction
- Problem to solve
- Why it's challenging
- What this guide covers

## The Challenge (200-300 words)
- Specific technical/business problem
- Why obvious approaches fail
- Requirements for solution

## Our Approach (300-500 words)
- Solution architecture/methodology
- Why this works
- Trade-offs considered

## Implementation (400-800 words)
- Step-by-step process
- Technical details (code, architecture)
- Pitfalls to avoid
- Time/resource requirements

## Results (200-300 words)
- Metrics from implementation
- Lessons learned
- What we'd do differently

## Conclusion
- Key takeaways
- When to use this approach
- Related resources

Case Study

Pattern: Customer → Problem → Solution → Results

Length: 1,000-1,500 words

Structure:

# {Customer Name}: {Result Achieved}

## Customer Overview (100-150 words)
- Company name, industry, size
- Business model
- Initial challenge

## The Problem (250-350 words)
- Specific pain point (quantified)
- Business impact (revenue, costs)
- Previous solutions attempted
- Why they needed change

## Solution (300-400 words)
- What they implemented
- Technical approach
- Implementation timeline
- Resources required

## Results (300-400 words)
- Metrics (before/after)
- Timeline to results
- Unexpected benefits
- Customer quote (if available)

## Technical Details (200-300 words)
- Architecture/integration approach
- Challenges overcome
- Why it worked

## Conclusion (100-150 words)
- Key success factors
- Applicability to others
- Soft link to product (if relevant)

Example (ElsaAI):

# ElsaAI Reduced Returns 38% with White-Label Fit Recommendations

## Customer Overview

ElsaAI is a luxury fashion marketplace with $200M GMV serving 250K monthly customers. Their curated selection emphasizes high-end designers and premium pricing—where brand consistency is critical.

They faced a 25% return rate on dresses, costing $3.2M annually in reverse logistics and lost revenue.

## The Problem

Luxury customers expect flawless brand experiences. ElsaAI's previous fit solution used a co-branded widget that broke their visual aesthetic:

- Widget displayed third-party branding
- Design didn't match site style
- Mobile experience was clunky
- Customers questioned legitimacy

Beyond aesthetics, fit accuracy was insufficient:
- Generic size charts: 68% accuracy
- Virtual try-on competitors: 72% accuracy
- Neither met the 85% threshold ElsaAI needed

At 25% return rates, $120 average order value, and 50K dress orders/year:
- Revenue impact: $1.5M (25% of $6M)
- Reverse logistics: $400K ($8 per return)
- Customer service: $200K (15 min per return @ $40/hr)
- Total annual cost: $2.1M

## Solution

ElsaAI implemented our white-label SDK with custom branding:

**Technical approach:**
- React SDK with custom styling
- ElsaAI's fonts, colors, design system
- Hosted on ElsaAI subdomain (fit.elsaai.com)
- Zero external branding

**Integration:**
- 2-week implementation
- 10 hours engineering time
- No infrastructure changes needed

**Fit accuracy:**
- AI trained on 10M+ body scans
- 94% accuracy on luxury dress category
- Personalized per customer

## Results

**30-day pilot (dress category only):**
- Return rate: 25% → 18% (28% reduction)
- Fit accuracy: 94% (vs 68% baseline)
- Customer satisfaction: 3.2 → 4.1 (27% improvement)
- Mobile conversion: +12% (cleaner UI)

**Annual projection:**
- Returns avoided: 3,500 (7% of 50K orders)
- Revenue retained: $420K
- Logistics saved: $280K
- Customer service saved: $105K
- Total annual savings: $805K

**ROI:**
- Implementation cost: $500K/year (white-label tier)
- Savings: $805K/year
- Net benefit: $305K (61% ROI)
- Payback: 7.4 months

## Technical Details

**Architecture:**
- SDK loaded asynchronously (no page speed impact)
- Body measurement API: <200ms latency
- Recommendation cache: 99.9% uptime
- Analytics dashboard: Real-time fit data

**Challenges overcome:**
1. **Brand consistency:** Custom CSS matching ElsaAI's design system exactly
2. **Mobile performance:** Lazy loading prevented slowdown
3. **Data privacy:** GDPR-compliant body measurement storage

**Why it worked:**
- White-label removed trust friction
- 94% accuracy exceeded luxury threshold
- Fast integration minimized engineering burden

## Conclusion

ElsaAI's success validates three insights:

1. **Luxury brands won't compromise on brand consistency** - Co-branded solutions are non-starters regardless of accuracy
2. **Fit accuracy >90% is table stakes** - Luxury customers expect near-perfect recommendations
3. **Fast implementation matters** - 2-week integration vs 3-month custom build

This approach works for fashion brands with:
- Strong brand identity (luxury, premium)
- High average order value (>$100)
- Technical team for integration (5-10 hours)

White-label SDK available for enterprise: Technical docs at {link}

Product Announcement

Pattern: What → Why → How → Who → When

Length: 400-700 words

Structure:

# {Feature/Product Name} Now Available

## What (100-150 words)
- Feature/product name
- One-sentence summary
- Primary benefit

## Why (150-200 words)
- Problem this solves
- Customer demand (from threads)
- Strategic context

## How (200-300 words)
- Technical approach
- Key capabilities
- Differentiation from alternatives

## Who (100-150 words)
- Target customers
- Use cases
- Requirements

## When & Pricing (50-100 words)
- Availability (now, beta, Q1 2025)
- Pricing tier
- How to access

## Technical Details (optional, 100-200 words)
- Architecture highlights
- Integration approach
- Performance specs

LinkedIn Company Post

Pattern: Insight → Analysis → Implication

Length: 300-600 words

Structure:

{Opening observation or data point}

{Analysis with supporting details}

{Implication or learning}

{Soft link to product/content if relevant}

Example:

We closed 5 enterprise fashion deals in Q4. All 5 chose white-label SDK over co-branded.

This wasn't about hiding our technology. It was about brand consistency being non-negotiable.

What we learned:

**Luxury brands ($100M+ GMV) prioritize brand consistency over vendor recognition**

They'll pay 3x more for white-label because:
- Co-branded widgets break visual aesthetic
- Customers question legitimacy of third-party branding
- Mobile experience needs seamless integration

One customer (luxury marketplace, $200M GMV) told us: "Our customers don't care who powers fit recommendations. They care that it looks like ElsaAI."

**Fast fashion brands ($10M-$50M GMV) prefer co-branded**

Opposite behavior:
- Third-party branding adds credibility ("powered by AI")
- Lower technical resources (easier implementation)
- Price-sensitive (white-label costs 3x)

**Takeaway:**
Customer segment dictates product packaging. Same technology, different positioning.

We're now offering both:
- White-label: Enterprise tier ($400K+)
- Co-branded: Growth tier ($50K+)

Technical docs: {link}

Email Newsletter

Pattern: Curated insights + Educational content

Length: 400-600 words

Structure:

Subject: {Specific value, not clickbait}

Preview: {Extend subject, don't repeat}

---

{Personal greeting}

{Opening: Why this matters now}

**Section 1: {Insight}** (150-200 words)
- Key learning or announcement
- Supporting data
- Link to full article

**Section 2: {Resource}** (150-200 words)
- Educational content
- Practical takeaway
- Link to guide/case study

**Section 3: {Update}** (100-150 words)
- Product update or industry news
- Brief context
- Link for more

---

{Closing line}

{Real person signature}
{Name, Title}

---

Quality Validation

Before outputting draft, verify:

Brand Voice Compliance

• Educational tone: Teaches, doesn't sell
• Technical depth: Includes architecture, data, methodology
• Data-driven: Specific numbers (not "many", "significant")
• Honest: Admits limitations, uncertainties
• No sales language: No buzzwords, CTAs, engagement tricks

Factual Accuracy

• All metrics from threads: No invented numbers
• Customer names verified: Check thread for public approval
• Technical details accurate: Match thread/Canvas exactly
• No confidential info: Remove proprietary details

Structure Compliance

• Pattern followed: Matches selected pattern guide
• Length appropriate: Within range for content type
• Complete sections: No "TODO" or missing parts
• Transitions smooth: Logical flow between sections

SEO Readiness (if blog)

• Keyword in title: Natural integration
• Keyword in first 100 words: Organic mention
• H2 subheadings: Descriptive, include variations
• Internal link opportunities: Note related content

---

Output Format

Draft File Structure

---
# Metadata (for tracking)
title: "{Draft title}"
content_type: "{blog|case-study|announcement|linkedin|email}"
target_keyword: "{primary keyword}"
source_thread: "{thread path}"
pillar: "{content pillar}"
created: "{date}"
status: "draft"
word_count: {count}
---

# {Title}

{Full content following pattern}

---

## Editor Notes

**For human review:**
- [ ] Verify customer name approval (ElsaAI - check with legal?)
- [ ] Confirm metrics accuracy (25% → 18% return rate)
- [ ] Add internal links (related articles on return reduction)
- [ ] Review technical depth (enough for credibility?)

**Potential improvements:**
- Add quote from ElsaAI customer (if available)
- Include architecture diagram (if exists)
- Link to technical docs (where hosted?)

Save Location

Drafts awaiting review:

threads/marketing/campaigns/{campaign-slug}/5-actions/drafts/{piece}-draft.md

Notification in ops/today.md:

## Content Drafts Ready for Review

1. **Case Study: ElsaAI White-Label Success**
   - Type: Blog article (1,450 words)
   - Location: threads/marketing/campaigns/luxury-validation-nov-2024/5-actions/drafts/elsaai-case-study-draft.md
   - Keyword: "white-label SDK"
   - Action: Review for accuracy, approve for SEO optimization

---

Edge Cases

Insufficient Source Material

If thread lacks details:

• Flag: "Insufficient data for {content_type}"
• Request: Human provides additional context
• Or: Suggest alternative content type (announcement vs case study)

Confidential Information

If thread contains confidential data:

• Anonymize: Remove customer names, specific metrics
• Generalize: "A luxury fashion brand" vs "ElsaAI"
• Or: Flag for human review before proceeding

Customer Approval Required

If using customer name/data:

• Flag: "Customer approval needed for public use"
• Provide draft but mark as "pending approval"
• Don't proceed to publication without confirmation

Multiple Threads as Source

If opportunity combines multiple threads:

• Read all source threads
• Synthesize learnings
• Note pattern: "Based on 8 deals across Q4"
• Higher confidence in conclusions

---

Success Metrics

Draft quality:

• Human approval rate: >80% (drafts accepted with minor edits)
• Revision rounds: <2 (before final approval)
• Factual errors: 0 (all data accurate)

Efficiency:

• Time to draft: <2 hours (AI generation)
• Human review time: <30 minutes (80% complete)
• Total time to publish: <1 day (draft → review → optimize → publish)

Content effectiveness (measured post-publication):

• Organic traffic: {sessions per article}
• Engagement: {time on page, scroll depth}
• Conversions: {demos requested per article}

---

Usage Example

Input:

topic: "How Enterprise Fashion Brands Use White-Label SDKs"
content_type: "case study"
target_keyword: "white-label SDK"
source_thread: "threads/sales/elsa-white-label/"
pillar: "Product capabilities"

Process:

1. Load thread:
   - Read Campaign: threads/marketing/campaigns/luxury-validation-nov-2024/4-decision.md
   - Read Source: threads/sales/elsa-white-label/6-learning.md
   - Extract: Customer (ElsaAI), Problem (returns), Solution (white-label), Results (38% reduction)

2. Load positioning:
   - Read: strategy/canvas/07-uvp.md
   - Extract: "AI-powered fit recommendations reduce returns"

3. Load voice:
   - Read: artifacts/marketing/narrative/brand-voice.md
   - Apply: Educational, technical, data-driven tone

4. Select pattern:
   - Load: {baseDir}/references/case-study-pattern.md
   - Structure: Customer → Problem → Solution → Results

5. Generate draft:
   - 1,450 words
   - Includes: Metrics (25% → 18%), Technical details (React SDK), ROI ($305K net benefit)
   - Voice: Educational (explains why luxury prefers white-label)
   - No CTA: Soft link to technical docs only

6. Validate:
   - ✓ Brand voice (educational, technical)
   - ✓ Factual (all numbers from thread)
   - ✓ Complete (all sections present)
   - ✓ SEO-ready (keyword in title, H2s)

7. Output:
   - Save: threads/marketing/campaigns/luxury-validation-nov-2024/5-actions/drafts/elsaai-case-study-draft.md
   - Flag: ops/today.md for human review

---

Remember

Content generation is:

• Transforming thread learning into educational content
• Applying brand voice (authoritative, not promotional)
• Following patterns (structure guides, not templates)
• Creating 80% complete drafts (human refines last 20%)

Content generation is NOT:

• Writing sales pitches disguised as content
• Inventing data or metrics
• Auto-publishing without human review
• Following rigid templates verbatim

Success = Educational content that builds authority and converts organically.