How this site is built

Step one: who qualifies

A restaurant gets onto a DineHub101 list when at least three independent published guides have named it. Three is the whole bar. We do not eat at these places, we do not score them ourselves, and we have no opinion about your dinner.

Why three? Because one list is a writer's taste, two lists can be two writers copying each other, and three separate outlets converging on the same restaurant is a signal that survives most kinds of nonsense.

Seoul, Taipei and Shanghai are the exception, and the bar there is two guides, not three. Not because those cities matter less. The lists that decide a restaurant's reputation in Seoul, Taipei and Shanghai — Blue Ribbon Survey, 500盤, the Dianping 必吃榜 — are published in ways that resist being read automatically, so far fewer of them reach us in a form we can check name by name. Agreement between three guides we can actually read is therefore structurally rarer there, and holding those three cities to the same three-guide bar would not have produced a stricter list — it would have produced a much shorter and more arbitrary one, made of whichever restaurants happened to be legible to us.

So two independent guides get a restaurant onto the Seoul, Taipei and Shanghai lists, and 282 of the restaurants on this site are there on two. Everything after that step is identical. The ranking formula, the weighting, the check date and the refusal to take money are the same in all ten cities; the only thing that changes is how many guides had to name a place before it was allowed to be ranked at all. Each venue page says how many named it, so you can hold any restaurant here to whichever bar you prefer.

"Independent" is doing real work in that sentence. We count the guides each city's own diners actually use, not just the English-language ones:

• Japan

  Tabelog, the Michelin Guide, and the restaurant press that Tokyo and Osaka argue about — plus long-running local awards.

• Mainland China

  Black Pearl (黑珍珠) and the Dianping 必吃榜, which is what people in Beijing and Shanghai genuinely consult.

• Korea

  Blue Ribbon Survey, the Michelin Guide Seoul, and Korean food media.

• Taiwan

  500盤, the Michelin Guide Taipei, and Taiwanese food press.

• South-East Asia

  Michelin for Bangkok, Singapore and Kuala Lumpur, alongside the local and expat food media that cover them week to week.

• Everywhere

  International food media — the travel and restaurant desks whose lists get read outside the country they describe.

Every restaurant on this site was named by two or more of those, and most by a good deal more: 491 of them appeared on four or more, and 113 appeared on six or more.

Step two: the order

Once a restaurant qualifies, we stop using the guides entirely. The ranking comes from one place: its Google rating, weighted by how many people left it.

A raw star rating is a bad way to sort a list. Five stars from nine people is not better than 4.6 from nine thousand, but a naive sort puts it on top. So we pull thin ratings toward the average of their peers, and let heavily-reviewed ratings stand on their own. The formula is:

That is the entire ranking system. There is no editorial thumb, no hand-placed favourite, no boost for restaurants that took our call.

What it looks like in practice

When we checked

Every rating and review count on this site was read on 24 August 2026. They are a photograph, not a live feed. Ratings drift slowly and review counts only climb, so the order changes little week to week — but a number here can be a few months stale, and we would rather say so than pretend to a freshness we do not have.

Nobody can pay

No restaurant paid to appear on this site, and no restaurant can. There is no advertising, no affiliate booking, no sponsored row, no "featured" tier, and no PR agency with our email address. A restaurant cannot buy its way onto a list, up a list, or off one.

The buttons on each restaurant's page link straight out to Google Maps, to its Google reviews, and to the restaurant's own website — and DineHub101 earns nothing from any of them. They are plain links, not affiliate or referral links, we take no commission on a booking, and we are not paid for the click.

If we ever take money from anyone connected to a restaurant, it will be written at the top of the page it affects, in the same size type as everything else.

What we count

This site loads nothing from anybody else's server, sets no cookies and runs no third-party analytics. It does count its own page views. Each one records the page, the language it was read in, and — when a browser sends one — the bare hostname the visit arrived from, so a search engine can be told apart from a link on somebody's forum. That is the whole record: no IP address, no browser or device details, no identifier of any kind, nothing that could be tied back to a person or joined up between visits, and nothing shared with anyone.

Clicks on the buttons that leave this site are counted the same way — a restaurant, a city and which button, never a visitor. Both counts exist so we can tell whether anybody is reading this, without following anybody around to find out.

What this is not

It is a shortlist, not a census

This is important, so it is not in small print: DineHub101 is a curated shortlist, not an exhaustive survey. There are tens of thousands of restaurants in these ten cities and thousands of very good ones. Each list here is about a hundred names long. A restaurant is missing here mainly because the guides we read had not named it enough times when we looked — not because it is bad, and not because we judged it. Absence from this site means nothing at all about a restaurant.

The bias runs in a specific direction, too. Guides over-cover restaurants that are easy for writers to visit and easy to write about, which means neighbourhood places, stalls and family kitchens are under-represented everywhere on this site. We know. We have not solved it.

One list, and cuisines are not rated alike

Each city page is a single ranked list running across every cuisine at once, and that hides something you should know about: Google ratings are not handed out evenly across kinds of food. In every one of our ten cities, the cheap, busy, everyday cuisine scores lower than the expensive imported one — usually by a wide margin.

Tokyo ramen shops average 4.23 on Google. Tokyo washoku restaurants average 4.55. Singapore's own cooking averages 4.02 while its Japanese restaurants average 4.59. Cantonese sits at the bottom of the Hong Kong table; Taiwanese sits at the bottom of the Taipei one.

Some of that gap is real. Most of it is not. A cheap bowl eaten in nine minutes at a counter collects impatient one-star reviews that a three-hour tasting menu never does, and diners grade against what they paid and what they expected. The effect is systematic, and because we sort the whole city on one number, it pushes each city's own everyday cooking down the page and lifts its French and Italian restaurants up. Only 2 of Tokyo's 33 ramen shops reach that page's top 101, while every one of its 23 Italian restaurants does.

We have not corrected for it, because any correction would be us putting a thumb on the scale and we would rather show you the raw thing and tell you what is wrong with it. This is what the cuisine filters are for. Tap Ramen on the Tokyo page and you get the ramen ranking, judged against ramen — which is the comparison that actually means something. The city-wide order is a starting point, not the answer.

Beijing and Shanghai are visitor picks, not a market verdict

Google is blocked in mainland China. Chinese diners do not use it, so a Beijing restaurant's Google rating is not the city's opinion — it is the opinion of the handful of foreign visitors who happened to open the app. The median Beijing restaurant on our list has around ten Google reviews. Shanghai is not much better.

So the two Chinese lists are built differently in the half that matters. The qualifying step is strong: Black Pearl and the 必吃榜 are what locals genuinely use, and the restaurants that appear are the right restaurants. The ordering step is weak, because the numbers behind it are thin. The weighting formula does what it can — a 5.0 from two reviewers gets pulled almost all the way back to the peer average — but no formula rescues data that thin.

Read Beijing and Shanghai as a visitor's shortlist of places the Chinese guides rate, in a rough order. Do not read them as a verdict from those cities' own diners. We label them that way on every page they appear on, and we would rather publish them with the warning attached than either drop two great food cities or quietly pretend the numbers are as good as Tokyo's.

Some restaurants have no Google listing at all

90 restaurants on this site — have no Google listing we could match with confidence — either none exists, or the closest match turned out to be a different business when we checked. They have no verified rating, so they get no rank. They sit at the bottom of their city page, because they earned their place through the guides and deleting them would quietly flatter our own coverage.

Matching to Google is imperfect

To read a rating we have to match a restaurant to a Google listing, and matching across scripts, romanisations, branches and renamings is genuinely hard. We match on name, address and neighbourhood, and we get some of them wrong. Where a listing looks like a plausible-but-different business we drop it rather than guess — but some bad matches certainly survive, most likely in the cities where the listings are thinnest.

Tell us when we are wrong

Wrong rating, wrong neighbourhood, closed restaurant, a guide we should be reading, or a match to the wrong Google listing — we would rather know. Corrections are the only kind of feedback that can change what appears on this site, which is exactly how we want it.

The short version

Three independent guides get you on the list, or two in Seoul, Taipei and Shanghai. Your diners decide where you sit. Money changes nothing. Two of the ten cities have data we do not trust and we say so on the page. That is all of it.