Stage 2: COMPOUND

| Path | Core Insight | Adopt | Reject | |------|-------------|-------|--------| | A: Full Local | Zero-cloud voice ordering on Jetson is architecturally clean | Queue analytics via depth centroid tracking, Piper TTS, cart state machine port | GPU contention risk, LLM-for-NLU overkill, memory pressure | | B: Companion iPad | 100% BWB code reuse, zero technical risk | V0.5 pilot strategy, WebSocket relay between LUME and companion device | Two-device proposition permanently (acceptable for V0.5, not V1) | | C: Cloud STT | Best accuracy in noisy environments, zero GPU impact | Cloud STT as primary with local Whisper fallback, BWB NLU patterns as domain expert | Single WiFi dependency without fallback | | D: Entertainment-First | The queue IS the product, order overlay in visual experience | Entertainment-first positioning, order confirmation integrated into visuals, $99 Venue tier | Weak standalone commerce, dependency on external POS for payment | | E: Multi-Vertical | Platform with swappable vocabularies and zone definitions | Plugin architecture designed in, extension points for future verticals | Multi-vertical GTM now (one beachhead first) | | F: Analytics-as-Service | Depth analytics is a proven $2.1B market, privacy-by-design | Analytics as enterprise selling surface, GDPR/CCPA-compliant depth-only sensing | Analytics alone doesn't justify $1,299, must demo entertainment | ## Compound Step 1: ESTABLISH GROUND TRUTH (The Product Identity) *Starting from scratch. No inheritance.* LUME Commerce is NOT a POS system with visual effects. It is NOT an analytics sensor with a pretty display. It is an **experiential commerce platform** where the entertainment experience IS the commercial engine. 1. **ENTERTAINMENT** (what draws them in): Depth-reactive visuals that make the queue magical. Customers interact with fluid light while waiting. Auto-captured content clips make every visit shareable. This is what the shop owner sees in the demo and says "I want that." 2. **INTELLIGENCE** (what keeps them paying): Privacy-preserving depth analytics that count bodies, measure wait times, track zones, and predict rush hours. The shop owner logs into a dashboard and sees data they've never had before. This justifies the monthly subscription after the novelty fades.