Emberline Hospitality

Mobile application

The challenge

Emberline's kitchens ran four screens: the POS, two delivery-platform tablets, and a printer for phone orders. Nothing sequenced across them, so a kitchen at peak was reconciling four independent queues by eye. Ticket times had crept up 22% over two years as delivery volume grew, and order errors attributed to the kitchen were rising in step. The venue Wi-Fi, in a majority of locations, was genuinely unreliable.

Research

• Peak-service observation across twelve venues spanning all five brands
• Analysis of 90 days of ticket-time data segmented by channel mix
• Wi-Fi reliability measurement at 40 representative venues
• Kitchen staff interviews on where the current setup failed them

The solution

Every channel — POS, three delivery platforms, phone, and first-party web and app — normalizes into a single order stream with consistent structure and timing metadata. The kitchen display sequences the combined queue by required fire time rather than arrival time, which is what the kitchen was approximating manually. Critically, the display is offline-first: it holds its full queue locally and continues operating through network loss, syncing state when connectivity returns. Given the Wi-Fi measurements, anything else would have failed in the majority of venues.

UX decisions

Features

• Unified order ingestion across seven channel types
• Fire-time sequencing across the combined queue
• Offline-first kitchen display with deterministic sync
• Multi-brand menu management with per-venue availability
• First-party ordering with integrated loyalty
• Course timing and expo coordination
• Venue analytics on ticket time, margin, and channel mix

Technology

Architecture

Channel adapters normalize inbound orders into a canonical schema published to Kafka. A Node service maintains per-venue queue state in PostgreSQL with Redis for hot queue reads. The React Native kitchen display holds a local SQLite queue as its working source of truth and reconciles through an append-only sync log — kitchen state changes always win over server state, since the kitchen is the physical reality. Menu configuration pushes down per venue with availability overrides that survive disconnection.

Results

Peak ticket time fell 31%. Kitchen-attributed order errors dropped 58%. First-party ordering grew from 19% to 44% of volume as the direct channel became genuinely faster than the delivery apps — worth substantially more per order given platform commission. All 180 venues went live over six weeks with no service-day incidents.

Lessons learned

1. 01Offline-first was not a resilience nicety here; the Wi-Fi survey made it a hard requirement. Measuring the venues before designing saved us from shipping something that worked in the office and failed in the field.
2. 02Removing every confirmation dialog felt reckless in review and was clearly correct in service. Undo is a better safety mechanism than confirmation when hands are busy.
3. 03The direct-channel growth was an unplanned second-order effect. Making first-party ordering faster than the aggregators shifted mix more than any marketing spend had.