Rfp submission

RFP Submission

Analyze any RFP (media, creative, CRM, full-service, performance) and generate a structured 8-chapter response that demonstrates strategic understanding, operational readiness, measurement rigor, and commercial clarity. Two-phase delivery: outline first (user approves), then full PPTX generation.

How This Skill Orchestrates

This skill is unique in the OS — it's the pre-sale engine that activates BEFORE the client is won. It uses the same OS skills that will operate the account, demonstrating them as capabilities during the pitch.

1. Script execution (scripts/rfp_analyzer.py): Classifies RFP type (5 types: media/creative/CRM/full-service/performance), maps requirements to the 8-chapter structure, generates section-level outline with slide estimates, builds research query plan, identifies MCP data sources, maps which OS skills power each chapter

2. Reference lookup (references/rfp_response_framework.md): 8-chapter deep guidance with section-level instructions, win theme methodology, JBP/AVB compensation models (3 AVB tiers + 6 compensation types), benchmark research protocol, pitch anti-patterns

3. Web search (Claude — CRITICAL): Before building chapters 2-3, search for client earnings, Kantar category data, Nielsen SOV, industry benchmarks, competitive landscape. This research IS the strategic differentiation

4. MCP tools (Claude): Google Drive for internal credentials/templates, Supermetrics for historical campaign data, Adobe MCP for platform capabilities, S&P Global for financial data if client is public

5. Upstream OS skills (Claude): Activate based on scope type (Working / Non-Working / Both):

   Working Media scope activates:

   • media-routing-planner → Chapter 3 (channel allocation, Hill curves, budget optimization)
   • mmm-modeling → Chapter 3 (saturation curves, ROI by channel) + Chapter 6 (Bayesian MMM roadmap)
   • media-procurement-benchmark → Chapter 3 (rate validation, supplier scoring, working capital)

   Non-Working Media scope activates:

   • crm-journey-architect → Chapter 3 (lifecycle design, SFMC/AJO journeys)
   • audience-segmentation-brief → Chapter 2 (research, consumer insights) + Chapter 3 (RFM, CLV, identity graph)
   • creative-supply-planner → Chapter 3 (asset matrix, GenAI production, fatigue scheduling)

   Always activated (both scopes):

   • fte-capacity-sizing → Chapter 4 (70/30 staffing model)
   • measurement-incrementality → Chapter 6 (measurement framework, adapted per scope type)
   • scope-audit → Chapter 8 (commercial review)
   • client-memory-synthesizer → If re-pitching existing client

6. File generation: Phase 1 → inline outline artifact. Phase 2 → PPTX deck via pptx skill (white-key template for editing)

Model decision: Interpretation → document intelligence + Scope → classification. The script classifies the RFP type and maps to chapters. The LLM does the strategic heavy lifting: synthesizing research, framing insights, building the narrative. No ML — it's structured analysis + web research + narrative generation.

Quick Reference

Resource	Purpose	Usage
scripts/rfp_analyzer.py	RFP classifier (5 types), 8-chapter outline generator, research query builder, MCP source mapper, OS skill activator	python rfp_analyzer.py --input rfp.json --output analysis.json
references/rfp_response_framework.md	Chapter-by-chapter guidance, win theme methodology, JBP/AVB models, compensation structures, benchmark protocol, anti-patterns	Read before building any chapter

Two-Phase Delivery Process

Phase 1: Outline (immediate)

1. Upload/paste the RFP document
2. Claude runs rfp_analyzer.py → classifies type, generates outline
3. Claude searches web for initial research (client, industry, benchmarks)
4. Produces inline visual artifact: chapter-by-chapter outline with sections, slide count, research findings preview, OS skills that will power each chapter
5. User reviews, adjusts, approves the outline

Phase 2: Full PPTX (post-approval)

6. Claude generates full content per chapter using research + OS skills
7. Produces PPTX via pptx skill — white-key template (clean, editable)
8. User and team edit, add credentials, finalize numbers in Chapter 8
9. Agency reviews and submits

Process

Step 1 — Ingest the RFP

Upload the RFP document (PDF, DOCX, or paste text). Claude extracts: client name, industry, scope (working/non-working media), budget range, timeline, evaluation criteria, specific questions to answer.

Step 2 — Classify and Map

Run the analyzer to classify RFP type and generate the 8-chapter outline. Each chapter gets section-level guidance, slide estimate, and OS skill mapping.

Step 2.5 — Apply Scope Routing (Working vs Non-Working Media)

After classification, Claude MUST determine whether the RFP is Working Media, Non-Working Media, or Both, and adapt the chapter content and skill activation accordingly.

Working Media scope (paid media investment flows through the agency):

• Chapter 3 activates: media-routing-planner (channel allocation, Hill curves, SLSQP optimization), mmm-modeling (saturation curves, ROI by channel, Bayesian MMM), media-procurement-benchmark (rate negotiation, supplier scoring, working capital)
• Chapter 6 emphasizes: media attribution, incrementality testing, geo-lift design, iROAS vs platform ROAS
• Chapter 8 includes: media management fee (% of spend), billing/reconciliation, make-good terms, transparency commitments
• Research adds: CPM/CPC/CPA benchmarks by platform, competitive SOV, media inflation trends

Non-Working Media scope (data, CRM, creative, analytics, consulting — no media buying):

• Chapter 3 activates: crm-journey-architect (lifecycle design, SFMC/AJO), audience-segmentation-brief (RFM, CLV, identity graph), creative-supply-planner (asset matrix, GenAI production), data strategy and CDP architecture
• Chapter 3 does NOT activate: media-routing-planner, mmm-modeling, media-procurement-benchmark — these are irrelevant when there's no media buying in scope
• Chapter 6 emphasizes: CRM incrementality (holdout tests), creative performance measurement, data quality metrics, journey attribution, not media attribution
• Chapter 8 includes: FTE-based retainer, project fees, deliverable-based pricing — no media management % because there's no media to manage
• Research adds: email/CRM benchmarks (open rate, CVR), creative production benchmarks, data maturity frameworks, platform licensing costs

Both (full-service) — all skills activate. Chapters 3, 6, and 8 get the full depth across media AND non-working disciplines.

How Claude applies this:

1. Check scope_working_media and scope_non_working flags from the analyzer
2. If only Working Media → emphasize media skills, de-emphasize CRM/creative depth
3. If only Non-Working → suppress all media planning/buying content, lead with CRM/data/creative
4. If Both → full depth, but clearly separate Working vs Non-Working sections within each chapter
5. The chapter structure (1-8) stays the same regardless — what changes is the CONTENT within chapters 3, 6, and 8 Before writing anything, define 1-3 win themes: why this agency is the best fit for THIS client. Evidence-backed, not aspirational. The win themes thread through every chapter.

Step 4 — Research (web search + MCP)

Execute the research query plan:

• Client earnings/investor presentations (S&P Global MCP or web search)
• Kantar Worldpanel category data (web search)
• Nielsen Ad Intel competitive SOV (web search)
• Industry benchmarks by platform (Triple Whale, WordStream, Pixis)
• Competitive landscape analysis

Step 5 — Build Strategic Chapters (2-3)

Use research to construct: insights → hallazgos → palancas estratégicas → tactical execution. Each insight must be non-obvious, data-backed, and connected to a business opportunity.

Step 6 — Build Operating Model (Chapter 4)

Show the 70/30 AI model. Show the OS skill chain. Show how 3-4 senior humans + Claude operate at the output level of 12-15 FTE. This IS the pitch differentiator.

Step 7 — Build Measurement Framework (Chapter 6)

Use the 5-level evidence hierarchy from measurement-incrementality. Show top-down (business → marketing → campaign) and bottom-up (delivery → incrementality) measurement architecture.

Step 8 — Frame Commercial (Chapter 8)

Provide the staffing framework (roles × dedication × rate tier), fee model options, and JBP/AVB structure. User fills the actual numbers.

Step 9 — Generate Outline Artifact

Produce the inline visual outline for review.

Step 10 — Generate PPTX (after approval)

On user approval, generate the full PPTX deck using the pptx skill.

Output Format

Phase 1: Outline Artifact (Visualizer)

Render as inline HTML widget:

• RFP Classification badge — type, confidence, emphasis
• Win themes — 1-3 themes in accent cards
• 8-chapter outline — expandable sections showing: title, section names, slide count, OS skills used, research status
• Research findings preview — key data points found via web search
• MCP data sources — connected sources with purpose
• Total slides estimate — target deck size
• sendPrompt() buttons: "Approve outline — generate full PPTX" → triggers Phase 2, "Adjust Chapter [N]" → lets user modify

Phase 2: PPTX Deck

White-key template via pptx skill:

• Clean, professional layout (not overly designed — your team will brand it)
• Content populated from research + OS skills + framework guidance
• Chapter 8 commercial sections as framework tables (user fills numbers)
• Speaker notes with talking points per slide
• ~52 slides total (adjust based on RFP type)

Markdown Summary

## 📋 RFP RESPONSE OUTLINE — [Client] — [RFP Type]

### Classification
Type: [media/creative/CRM/full-service/performance]
Win themes: [1-3 themes]

### Research summary
[Key findings from web search — market, category, competitive]

### Chapter outline
| Ch | Title | Sections | Slides | OS skills | Status |
| 1 | Credenciales | 4 sections | 4 | — | Template |
| 2 | Planteamiento estratégico | 3 sections | 8 | audience-seg, measurement | Research needed |
...

### Commercial framework (Chapter 8)
[Staffing grid template + fee model options + JBP/AVB structure]

### Next steps
1. Review and approve outline
2. Claude generates full PPTX
3. Team edits credentials, adds numbers, brands the deck
4. Internal review → submit

Examples

Example 1 — Media RFP: User uploads GlobalBev media RFP for Mexico. → Classified: Media (92% confidence). 52 slides. Research: Kantar shows GlobalBev at 62% penetration, Nielsen SOV at 35% in non-alc beverages. Win themes: "LATAM media expertise" + "AI-native optimization." Chapter 3 shows media routing with Hill curves. Chapter 6 shows incrementality roadmap.

Example 2 — Full-service pitch: User: "We're pitching AcmeAuto's full digital scope — media, CRM, creative, data." → Classified: Full-service. 58 slides (all 8 chapters deep). Research: AcmeAuto MX earnings show 15% digital investment increase YoY. Win themes: "Automotive category depth" + "SFMC + AJO dual expertise" + "70/30 model = 3× margin efficiency."

Example 3 — CRM RFP (Non-Working): User pastes GlobalCosmetics CRM RFP. → Classified: CRM (85%). Scope: Non-Working Media only. 44 slides. Media skills suppressed — no media-routing, no mmm-modeling, no procurement. Chapter 3 leads with crm-journey-architect (SFMC lifecycle) + audience-segmentation-brief (RFM + CLV). Chapter 6 measurement = CRM incrementality (holdout test), not media attribution. Chapter 8 = FTE retainer, no media management %. Win theme: "Lifecycle architecture on SFMC with identity graph."

Example 4 — Data/Analytics RFP (Non-Working): User: "RFP for audience data strategy, CDP implementation, and analytics reporting. No media buying." → Classified: Performance (70%). Scope: Non-Working Media. Chapter 3 activates audience-segmentation-brief (identity graph, 1P/2P/3P data, consent framework). Media skills fully suppressed. Chapter 7 gets extra depth (data architecture, CDP, data clean rooms). Chapter 8 = project-based + retainer hybrid.

OS Routing

This skill sits BEFORE the commercial cycle:

RFP arrives → rfp-submission (analyze, research, propose)
  → [WIN] → client-memory-synthesizer (create tenant context)
    → intake-router → scope-audit → ... (operating cycle begins)
  → [LOSE] → learnings captured in memory for next pitch

Skill Chaining

Direction	Skill	Connection	Scope type
Upstream	agency-request-intake-router	Routes "new business" requests here	All
Downstream (during pitch)	audience-segmentation-brief	Powers Chapter 2 research + Chapter 3 segmentation	Non-Working + Both
Downstream (during pitch)	media-routing-planner	Powers Chapter 3 channel allocation + optimization	Working Media + Both
Downstream (during pitch)	mmm-modeling	Powers Chapter 3 saturation/ROI curves + Chapter 6 Bayesian MMM	Working Media + Both
Downstream (during pitch)	media-procurement-benchmark	Powers Chapter 3 rate validation + Chapter 8 supplier terms	Working Media + Both
Downstream (during pitch)	crm-journey-architect	Powers Chapter 3 lifecycle design (SFMC/AJO)	Non-Working + Both
Downstream (during pitch)	creative-supply-planner	Powers Chapter 3 asset matrix + GenAI production	Non-Working + Both
Downstream (during pitch)	fte-capacity-sizing	Powers Chapter 4 staffing model	All
Downstream (during pitch)	measurement-incrementality	Powers Chapter 6 measurement framework	All
Downstream (during pitch)	scope-audit	Validates Chapter 8 scope/commercial	All
Downstream (post-win)	client-memory-synthesizer	Creates tenant context for new client	All
Downstream (post-win)	ALL operating skills	Begin service delivery	All