Dei Martinez Elurbe

Some questions don't need your data. They need the data sitting in public view: prices, listings, menus, reviews, registries, whatever the market leaves lying around. I collect it and build the model that turns it into an answer you can act on. Behind it: a PhD, ten years as a computational biologist, and the habit of asking what evidence has to survive before anyone acts on it.

This summer I measured Amsterdam restaurant prices from nothing but the public web: 900 scraped menus, 33,000 dishes, one model. The tourist trap is real and costs about 85 cents on a €20 main; what decides your bill is what you order, not where you sit. Full study: deita.eu/amsterdam-menus

The restaurant part is incidental. The same market research questions show up everywhere: what does the market actually charge, and where do we sit? Is the premium we suspect real, and how big? If the raw material is public, a serious answer is usually buildable. I work with AI as core tooling; it's how one person covers 900 restaurants in a summer.

Half the craft is building the instrument the question needs. For the menus that was the bitterbal, a snack sold at the same spec everywhere, so any price gap is pure location. Your problem will need a different instrument; building it is the part I'm good at. The other half is knowing when to stop. The study set out to grade every restaurant as a deal or a premium. The numbers couldn't carry a verdict that specific, so I killed it and built the question they could answer: the going rate for every dish in town. I'd do the same with yours: tell you what won't hold, and find what will.

An engagement is a scoped, fixed-price study: a price benchmark, a competitive analysis, a market scan. You bring the question, we agree scope and price, you get an analysis you can act on and, if it serves you, publish.

Extra depth in food and hospitality: I grew up in a family of restaurateurs and run deita (pricing, menus, demand forecasting). But the method travels.