Pre-Sales Estimation Template

An AI-powered pre-sales estimation workflow for analyzing TOR/RFP/SOW documents, drafting clarifying questions, reviewing estimates, and producing gap analysis — all driven by Claude Code.

Quick Start

# From this directory
./init.sh <client-name> [tech-stack]

# Examples
./init.sh Ferrellgas-BlueRhino drupal
./init.sh Acme-Corp drupal-nextjs
./init.sh BigRetail nextjs
./init.sh MyClient react

This creates a new engagement directory as a sibling to _template/, initializes git, and activates the appropriate technology overlay.

First time? See SETUP.md for prerequisites and installation.

Supported Tech Stacks

Stack	Overlay	Default?
Drupal 10/11	drupal_estimation	Yes
Drupal + Next.js (Headless)	drupal_nextjs_estimation	No
Next.js	nextjs_estimation	No
React	react_estimation	No

Each overlay activates tech-specific CARL rules that guide estimation quality (e.g., Drupal overlay enforces checks for contrib vs custom modules, CMI, decoupled overhead; Drupal + Next.js headless overlay adds rules for API design, JSON:API/GraphQL selection, SSR/SSG strategy, and decoupled architecture patterns).

Workflow

The template uses a 6-phase workflow, each triggered by placing documents in the corresponding directory and prompting Claude Code:

Phase 0: Customer Research      →  Place TOR in tor/, ask Claude to research
Phase 1: TOR Analysis           →  Ask Claude to analyze TOR and generate questions
Phase 1A: Optimistic Estimate   →  No-response path: /optimistic-estimate → /tech-proposal
Phase 2: Response Integration   →  Place customer answers in responses_qna/
Phase 3: Estimate Review        →  Place estimates in estimates/
Phase 4: Gap Analysis           →  Ask Claude to generate gap analysis
Phase 5: Knowledge Capture      →  Ask Claude to store learnings (post-engagement)

Two paths after Phase 1:

• Standard path (Phase 2→3→4): Customer responds to questions, estimates reviewed and validated
• No-response path (Phase 1A): Customer Q&A not received — generate optimistic, assumption-heavy estimates and a Technical Proposal directly

Phase Prompts

Phase	Prompt
0	"Research the customer and audit their existing site based on the TOR in tor/"
1	"Analyze the TOR in tor/ and generate clarifying questions"
1A	/optimistic-estimate then /tech-proposal
2	"Customer responses are in responses_qna/. Analyze them against the TOR and original questions"
3	"Estimates are in estimates/. Review them against requirements and responses"
4	"Generate the gap analysis and revised estimates"
5	"Capture learnings from this engagement for future reference"

Star-Commands

Available in any phase via CARL:

Command	What it does
*checklist	Runs the full estimation gap checklist against all artefacts
*recap	Summarizes engagement state: phases done, open questions, gaps, confidence
*benchmark <query>	Looks up effort benchmarks from memory and benchmarks/
*optimistic	Generate optimistic estimate (no-response path)
*proposal	Generate Technical Proposal Document

Directory Structure

<client-name>/
├── CLAUDE.md                    # Workflow instructions (auto-populated by init.sh)
├── .claude/settings.json        # Plugin config
├── .carl/
│   ├── manifest                 # Active CARL domains
│   ├── presales                 # Core estimation rules (always on)
│   ├── commands                 # Star-commands
│   └── <tech>_estimation        # Tech-specific rules (activated by init.sh)
├── .claude/commands/             # Project-level slash commands
│   ├── optimistic-estimate.md   # /optimistic-estimate (Phase 1A)
│   └── tech-proposal.md         # /tech-proposal (Phase 1A)
├── tor/                         # Input: TOR/RFP/SOW documents
├── initial_questions/           # Output: Phase 1 clarifying questions
├── responses_qna/               # Input: Customer responses
├── estimates/                   # Input/Output: Estimate documents
├── estimation_template/         # QED42 Excel estimation template
│   └── QED42-Estimate_Template.xlsx
├── claude-artefacts/            # Output: All AI-generated analysis
├── scripts/                     # Utility scripts
│   └── populate-estimate-xlsx.py  # Populates Excel from markdown estimate
├── templates/                   # Output structure templates
│   ├── tor-assessment-template.md
│   ├── questions-template.md
│   ├── estimate-review-template.md
│   ├── gap-analysis-template.md
│   ├── optimistic-estimate-template.md
│   └── technical-proposal-template.md
└── benchmarks/                  # Effort reference ranges
    ├── general-effort-ranges.md
    ├── drupal-effort-ranges.md  # (if Drupal stack selected)
    └── drupal-nextjs-effort-ranges.md  # (if Drupal + Next.js headless selected)

Output Templates

Each phase produces artefacts following a consistent structure defined in templates/. This ensures every engagement produces comparable, reviewable documents with:

• TOR Assessment: Requirements table with clarity ratings, risk register, assumptions log
• Questions: Domain-grouped questions with requirement references and suggested options
• Estimate Review: Coverage matrix, orphan detection, effort reasonableness flags, missing categories checklist
• Gap Analysis: Full traceability matrix, gaps/orphans lists, revised effort suggestions, confidence scores
• Optimistic Estimate: Domain-grouped estimate with CR-boundary assumptions, question→assumption mapping, coverage checklist
• Technical Proposal: Client-facing 11-section proposal with architecture, solution approach, team composition, and phased scope boundaries

Benchmarks

The benchmarks/ directory contains reference effort ranges for calibrating estimates:

• general-effort-ranges.md — Tech-agnostic ranges (PM overhead, QA, DevOps, migration by volume, integration tiers)
• drupal-effort-ranges.md — Drupal-specific ranges (content types, custom modules, contrib config, theme dev, migration sources)
• drupal-nextjs-effort-ranges.md — Headless-specific ranges (API layer, Next.js frontend, decoupled auth, preview/draft mode, ISR/SSG overhead, GraphQL/JSON:API integration)

These are starting points. Update them with actuals from completed engagements via Phase 5 (Knowledge Capture).

Adding a New Tech Overlay

1. Create a new file in .carl/overlays/ following the naming pattern <tech>_estimation
2. Add rules using the format <TECH_UPPER>_ESTIMATION_RULE_<N>=rule text
3. Update init.sh:
   • Add the tech name to the VALID_TECHS array
   • Add a case to the TECH_NAME mapping
4. Optionally add a benchmarks/<tech>-effort-ranges.md file

License

Internal use. Adapt as needed for your organization.