Key Decisions - February 11, 2026

Decisions made during the APEX POC kickoff meeting with Bob, Kartik, and Sabrina.

1. Initiative Structure

Decision: Initiatives are directories, not single files.

Structure:

initiatives/<name>/
├── Initiative.md
├── meetings/
├── discovery/
├── experiments/
├── designs/
├── tickets/
├── release-notes/
└── docs/

Rationale: Allows all related artifacts to live together, enables natural organization as initiative grows.

2. POD as Metadata

Decision: Team (POD) assignment is frontmatter metadata, not folder structure.

pod: Revenue-Optimization

Rationale: Teams change, initiatives move between teams. Metadata is easier to update than reorganizing files.

3. Repository Organization

Decision: Domain-centered product repositories.

Options Discussed: - Repo per microservice (rejected - 147 repos is too many) - Single monorepo (rejected - too broad, legacy complications) - Product surface repos (selected)

Proposed Repos: - product-pricing - Revenue pricing - product-forecasting - Demand forecasting - product-reporting - Analytics & dashboards

Alternative: Customer-facing names (e.g., "Game Changer")

Action: Kartik/Sabrina to finalize naming convention.

4. Branching Strategy

Decision: Start without PR workflow.

Rationale: - Reduce friction for PM adoption - Work in different directories = trivial merging - Add PRs later if conflicts arise

Future: Add PR workflow for design review checkpoints.

5. Technology Stack for New Projects

Decision: Greenfield stack for Project Searchlight.

Component	Choice
Frontend	Vercel
Auth	Clerk
Data	Lighthouse API
Backend	New (not monolith)

Rationale: Searchlight uses only Lighthouse data. Perfect test case for new stack without monolith dependency.

6. PRD Generation

Decision: PRDs are auto-generated from initiative context.

Inputs: - Initiative document - Meeting notes - Discovery artifacts - Experiment results - Designs

Rationale: Reduces manual work, ensures consistency, provides structured input for engineering AI agents.

7. Ticket Creation

Decision: Engineers create tickets from PRD, not PMs.

Rationale: Engineers understand technical breakdown better. PM expertise is problem/solution space, not task decomposition.

8. Release Notes Integration

Decision: Release notes are part of initiative directory.

Auto-generation sources: - PRD context (the "why") - Code changes - Experiment learnings

9. Pilot Project Selection

Project	Stack	Rationale
Project Searchlight	Greenfield	Pure Lighthouse data, no monolith
Demand360	Existing	Tests migration path
Tour Operator	Greenfield	Already has artifacts

Open Questions

Repo naming convention - Product surface vs customer-facing names?
Sub-module linking - How to link initiatives to code repos?
Egress costs - Vercel + AWS backend data transfer costs?
Contract requirements - Cloud vendor restrictions?

Next Steps

Wednesday kickoff meeting (Feb 19) with Shiv, Charles
Bob to polish Cursor skills
Pull Tour Operator artifacts from Confluence
Test PRD generation with Searchlight context

Key Decisions from POC Kickoff