Why a data science company runs a ghost kitchen
Decision Science Corp is an applied data science shop. We are not a restaurant company. So why does our dev blog spend thousands of words on DoorDash, ghost kitchens, Google Ads for restaurants, and Meta creative tests?
Because we run a field experiment: a real ghost-kitchen chain called Empanada Empire — handcrafted empanadas, Richardson TX, delivery and pickup, no dining room. It is not a side project we pretend to operate. It is the fastest honest way we have found to study marketplace funnels, menu and POS integration, off-platform ads, and operational feedback loops under real fees, real reviews, and real late-night fulfillment constraints.
Why a ghost kitchen is a useful lab
Traditional restaurants mix signal: foot traffic, word of mouth, local reputation, and digital. A ghost kitchen strips that down. Every order is digital. Discovery is apps and ads. Conversion is listing photos, price, ETA, and star count. Retention is reorder rate and promo response. That is close to the systems we build for clients — except we own the P&L and the blame when something breaks.
- Marketplaces (DoorDash and peers) — cold start, sponsored placement, BOGO economics, review velocity
- First-party ordering — empanadaempire.us on our multihost stack, shared menu DB with Square and delivery channels via Kitchen POS
- Paid social and search — Google Ads, Meta, creative testing, attribution into orders
- Ops and software — one menu database, config-driven radius and copy, agents and internal tools on the same Tasks API we ship for clients
What we are not doing
We are not publishing DoorDash tips because we want catering leads. We are not anonymously affiliate-farming restaurant content. We are documenting what we measured after doing the work — including the boring parts: spreadsheet payroll mistakes, certbot afternoons, and YouTube research marathons that become playbooks.
How to read the series
Posts that extend this experiment link back here. If you only care about infrastructure, read the Kitchen POS and multihost pieces. If you care about marketplace cold start, start with the DoorDash notes. If you care about paid acquisition off the apps, read the Google Ads and Meta summaries. Everything ties to the same live brand.
Where we are now
Empanada Empire is operational on production infrastructure: public ordering site, POS-backed menu, Texas ghost-kitchen address, and ongoing research corpora (DoorDash-native marketing, local Google Ads, paid social) that we synthesize into internal playbooks before we distill them on this blog. The food is real. The fees are real. The lessons are what we are selling when we say we extract signal from messy operational data.
Questions about the experiment or the stack: contact us here.