Advisory · Dei M. Elurbe, PhD · Amsterdam
Answers you can defend
I am a decision scientist for the companies around restaurants: POS platforms, field-sales tools, wholesalers, funds. They come to me when their data has more to say than it currently does, and when the numbers they put in front of clients, journalists and investors have to hold.
Snapshot · June 2026 · 32,405 dishes · 930 venues
The Amsterdam menus study
In July 2026 I published a study of Amsterdam's restaurant menus: a June snapshot of 32,405 dishes across 930 venues, about 4 in 10 of the city's restaurants. It is built only from public data, and all of it is public in return: the essay, the technical dossiers, the code, the data. The model has one job: refuse unfair comparisons. Never a steakhouse against a snackbar, never a dessert against a main.
Three of its findings, as the study states them:
Snapshot · June 2026 · 4,375 dishes · 528 venues
An ordinary dinner: €35–38, everywhere
bigger dots are areas with more venues · widest gap between areas: about €3
Burrata, a pasta, a cheesecake: €35 to €38, in every part of the city. No area breaks away by more than the noise. The dish decides the bill; the neighbourhood barely moves it.
methods →Snapshot · June 2026 · 34,996 dishes · 899 venues
Prices follow where tourists stay
+4.2% per step up in tourist exposure, 90% credible interval +1.8 to +6.6 · the dots are that one gradient mapped onto each borough's own exposure, not seven separate estimates
Dishes cost about 4.2% more per step up in tourist exposure (one standard deviation on the exposure index). The premium tracks where tourists sleep, not how close the landmarks are. Where it sits, not why.
methods →Snapshot · June 2026 · plain counts
Steakhouses crowd the centre
share of venues that are steakhouses, then the mirror: 5 of the city's 10 tagine kitchens sit in Nieuw-West
Steakhouses appear in the centre 2.6 times as often as in the rest of the city; Nieuw-West has none at all, and holds half the city's tagine kitchens. The same counts show where a concept is absent.
the selection story →Each card states its own cut of the June 2026 snapshot; the dossiers state every model's universe and interval conventions.
For companies that build on restaurant data
The questions it answers
The method travels: any market where menus, prices and venues can be observed, any claim that has to survive a skeptical reader. Four examples of what it hands you, depending on where you sit.
For field-sales platforms and distributors
What can your reps say in front of a venue?
A price reference your people can quote: what is typical for a dish, what counts as expensive, where a venue sits against its true peers. A rep who knows the market better than the venue does starts a different conversation. The dish price atlas is that reference, built for Amsterdam.
For companies that hold data
What could your data become?
POS systems, payment platforms and data providers sit on data their clients would pay to understand. I tell you which insight products yours could carry, which claims would survive a hostile read, and what it takes to turn a raw corpus into a dataset your team can build on.
For wholesalers and brands
Where is the white space?
Which concepts crowd which neighbourhoods, and where a category is missing entirely. Plain counts, so you know where to point a salesforce or launch a product. The steakhouse card above is that answer for one concept.
For anyone about to act on a number
Will this finding survive scrutiny?
Before you publish a number or build on one, I stress-test it the way I test my own published work: the threat, the test, the verdict, one line each. If it breaks, better it breaks with me.
How work starts
A rough question is enough
I do not pitch cold and I do not run outbound, so the first step is yours. There is no sales team behind this page, only the person who does the work, happy to think along before anything is agreed.
- Write me. Email or LinkedIn. No polished brief: two sentences about what you are trying to decide, or about what your data will not tell you, is plenty. Within a day I tell you whether I am the right person for it. If I am not, I say so.
- We talk. It costs nothing and commits you to nothing. Bring doubts rather than specifications; sorting them out together is what the call is for. I do not quote numbers on calls.
- You get a written proposal: what can be done and how, what cannot be done and why, a fixed scope, a fixed duration, and one fixed price. The work itself arrives in writing: dossiers, reference tables, protocols, and a readout call.
- You decide with the document in hand, and share it with whoever needs to see it.
Most engagements run two to six weeks, remote. If you use an NDA or a DPA, I sign yours.
The other half of the practice
deitalite
The menus study is one half of what I do. The other is deitalite: demand forecasting for restaurants. I designed it, I built it, and I run it every day.
Standing terms · the same for everyone
- Fixed fee. One number, in writing, before anything starts. No hourly billing, no overage. If the work takes longer than I estimated, that is my risk.
- One engagement at a time. Most run two to six weeks, part-time and remote. If I am mid-engagement when you write, I tell you when I can start.
- Payment. 50% on start, 50% on delivery. Invoiced from the Netherlands; VAT handled according to where your business is registered.
- What you receive. Written documents your team can build on: dossiers, reference tables, written protocols, and a readout call. No slideware.
- Your NDA or DPA, your environment. If you use one, I sign yours. If your data cannot leave your environment, a synthetic mirror with the same shape works.
- Confidentiality. Your data stays yours and is used only for your engagement. It is never resold, never passed on, and never used as a case study without your written permission.