I Counted Every Restaurant in 22 Countries. It Raised More Questions Than Answers.

For almost 30 years, my career has been about figuring out how to use data to explain things. As an engineer in the early days of data warehousing, I used it to help businesses understand their operations and financials. As a Bain consultant, to understand markets and support strategic decisions. At VoloMetrix, we were the first to start leveraging enterprise collaboration data as an entirely new source for understanding how organizations actually work. That one became a fascinating ten-year journey of going deeper and deeper into what was possible with such a large and varied dataset, and the extent to which it could explain incredibly complicated organizational systems. We were able to publish lots of very interesting case studies in HBR and in collaboration with academia on this.

Lately, I’ve become fascinated with restaurant data. It’s more relatable (we all eat out at least sometimes), and the deeper I get into it, the more interesting I find its connection to people and culture. Recently I came across data that shows full counts of food and drink establishments by country, and I’ve lost countless hours exploring it. Below I share some of the journey.

South Korea has over four times as many food and drink venues per capita as the United Kingdom. That sounds like it should mean something. It sounds like Koreans eat out constantly and the British cook at home. It sounds like Seoul is drowning in restaurants and London is a culinary desert. But, I’m not sure it means that at all.

I counted every rated restaurant, bar, pub, and cafe across 22 countries using the same data source, same filters, same methodology.

Here’s the chart I expected to be the punchline of this story:

At first glance, you’d assume South Korea simply eats out more than Britain. That was exactly the conclusion I expected this chart to deliver.

I thought maybe there would be some standard ratio of restaurant density to population. Nope, not at all. This chart is the beginning of the investigation, not the answer. The question, then, is why does this look so different across countries? I spent a week pulling threads and a picture began to emerge. Restaurant density by itself doesn’t tell you very much. It’s a mixture of tourism, venue design, affordability, local habits, and possibly other factors all blended into a single number. I got there by testing a series of hypotheses.

Theory 1: Maybe people in some countries just eat out more

Everyone eats about 21 meals a week. If just 5% of those happen at a restaurant, Portugal’s 62,000 venues each need to serve 24 meals a day to cover that demand. One decent lunch service would probably suffice. The UK’s venues need to serve 62. Southern European countries like Portugal can sustain their venue count even if residents rarely eat out. UK and US venues need to serve more people per seat to keep up.

Interesting. But it doesn’t explain the gap. It just restates it. I needed a better theory.

Theory 2: Maybe it’s the tourists

Greece gets 37.9 million tourists a year. Population: 10.4 million. Those tourists eat. Maybe countries with lots of restaurants just have lots of tourists.

So I sampled specific cities and islands within six countries, comparing tourist-heavy areas with places tourists rarely visit, and plotted restaurant density against tourism pressure:

Greece told a clear story. The islands have 24 restaurants per 1,000 people. Thessaloniki, where actual Greeks live and work, has 1.5. That’s below the UK national average.

The reason is straightforward. Tourists don’t usually have kitchens. They eat virtually every meal out. And in places like Mykonos, on a typical summer day, tourists outnumber residents by roughly six to one. Even conservative assumptions imply they generate the overwhelming majority of restaurant demand. Greece’s reputation as a restaurant culture is, by this measure, a tourism artifact.

Look at how dramatically the annual tourist-to-resident ratios vary (note: this is total annual throughput, not daily presence; average stays range from ~2 nights for a European city break to ~4-5 nights on a Greek island):

Mykonos: 280 tourists per resident (280!!!). Barcelona: 13. Columbus, Ohio: effectively zero.

I thought I’d found the answer.

Then Spain ruined it.

Barcelona receives 13 tourists per resident. Valladolid receives about 2. Their total venue density is nearly identical: 4.8 versus 4.7 per 1,000. The tourism ratio is 6x but the venue count barely moves. Spain’s density barely changes whether you look at a tourist city or a domestic one. Whatever is driving it, it isn’t visitors.

Tourism explains Greece. It doesn’t explain Spain.

Every time I thought I’d found the explanation, another country broke it.

Las Vegas broke it from the opposite direction.

The venue format problem

Vegas receives 57 tourists per resident, more tourism pressure than Venice or Florence. But it has only 1.9 restaurants per 1,000 people. Santorini, at 130 tourists per resident, has 24.4.

Same dynamic. Wildly different density. Why?

A Santorini taverna seats 25 people. A Vegas casino restaurant seats 300. But it’s not just seat count. American restaurants are engineered for throughput. Tipping culture creates a financial incentive for servers to turn tables fast. Three to four seatings per table per evening is normal in a busy US restaurant. In most of Southern Europe, a dinner table is yours for the evening. Getting a server to bring the bill is sometimes its own adventure.

Run the math: a 300-seat Vegas restaurant doing 3.5 turns serves about 1,050 covers a night. A 25-seat Santorini taverna doing 1.5 turns serves 37. Same industry, roughly 28x difference in output per venue. Restaurant density measures number of establishments, not meals. Two places can serve a similar volume of people and look completely different on a pure venue density chart due to capacity and throughput.

This probably explains a lot about why the US national density (2.3/1K) looks so low. American restaurants are just bigger, and they’re built to move people through faster.

Theory 3: Maybe it’s the price

I looked at the cost of a basic restaurant meal relative to average wages. The theory: if eating out is cheap, more people do it, more restaurants exist.

To make meal costs comparable across countries, I converted them to minutes of work. I took the average annual wage in each country (from the OECD, adjusted for purchasing power) and the cost of an inexpensive sit-down restaurant meal (from Numbeo’s crowdsourced database). Dividing one by the other gives a simple measure: how long does the average worker need to work to pay for lunch?

In Japan, a restaurant meal costs about 15 minutes of work. In the UK, 40. Japan has twice the venue density. The pattern holds at the extremes. (Greece shows 66 minutes, but that figure likely reflects tourist-area pricing rather than what a worker in Thessaloniki pays for lunch. The same tourism distortion, showing up in a different dataset.)

But Germany breaks it. Meals cost 28 minutes of work there. Relatively affordable. Yet Eurostat shows Germans spend just 4.1% of household budgets on restaurant meals, the lowest in Europe. Spaniards spend 12.7%.

Affordability creates the conditions for a dining-out culture. It doesn’t create the culture itself. Germany can afford to eat out. It chooses not to.

And Spain? Meals there cost 37 minutes of work. Mid-range. Not cheap. But, despite that, Spaniards spend three times as much of their income on eating out relative to Germans. Spain broke another model. Again.

So what does density actually measure?

Four factors, layered together, account for most of what I found.

Tourism sets the ceiling. The most restaurant-dense places on the planet are small tourism economies. Tourists eat out way more frequently than locals. When you look at places tourists don’t go, Greece drops from 6.0 to 1.5 venues per 1,000. Spain barely moves. That’s the test: does density survive when tourists aren’t there?

Venue format shapes the count. Santorini has 12x the restaurant density of Vegas despite lower tourist pressure. Small, slow-turn venues inflate the count; large, high-turn venues compress it. Density measures establishments, not meals served.

Affordability sets the floor. Where a meal costs under 20 minutes of work, eating out can be a daily default. Where it costs 40 minutes, it becomes an occasion.

Culture is the residual. After accounting for tourism, venue format, and affordability, Spain still has 2-3x the venue density of Germany and France. Spain spends 12.7% of household budgets on restaurant meals. Germany spends 4.1%. Whatever drives that difference, the data can measure its effect but can’t explain its cause.

The mystery I haven’t solved

Then I noticed something I couldn’t explain.

Houston.

At 0.4 restaurants per 1,000, it’s the lowest in my entire dataset. Lower than domestic Greece. Lower than Birmingham, England. American suburbs seem to organize eating differently from everywhere else. The car, the drive-through, the strip-mall chain might serve the same function as a European neighborhood bar, but they create a radically different landscape.

And then there’s Birmingham, England and Columbus, Ohio. Both sit at 0.5 restaurants per 1,000, on opposite sides of the Atlantic, with entirely different food cultures. Is there something about Anglophone car-suburb culture that creates a ceiling for restaurant density? Or is this just what happens when eating out costs 30-40 minutes of work?

I don’t know.

The biggest surprise from this research wasn’t that South Korea has more restaurants than Italy. It was discovering that “restaurant culture” isn’t one thing. Sometimes it’s tourism. Sometimes it’s economics. Sometimes it’s tiny tavernas instead of casino buffets. And sometimes it seems to be something the data can’t quite reach.

Spain, I’m looking at you.

If you have a theory, I’d love to hear it.