We don't list every venue in the city. Here's how we choose what gets in.

A friend visiting Shanghai once asked me how many restaurants there were in the city. I said I didn't know exactly — somewhere over 100,000. He asked how many were on Unight.

"About 500."

He looked at me like I'd done the math wrong.

I hadn't. The number was deliberate. It's still deliberate. There are around 500 venues on Unight in Shanghai not because we couldn't find more, but because most restaurants and bars in any city — including this one — are not actually worth your time. And the entire job of a curation platform is to know the difference.

This is how we decide what gets in.

The bar is "would I send a friend here?"

Every venue we consider goes through a single mental test before any spreadsheet or score: would I actually send a friend here?

Not would a tourist enjoy it. Not is it Instagrammable. Not does it have a celebrity following. Would I, knowing this friend's taste, knowing what tonight is for, send them here and feel confident they'd thank me afterward?

It sounds simple. It's the hardest filter we have. Most venues fail it for reasons that don't show up on Google Maps — service that's inconsistent, food that was great two years ago and isn't anymore, a room that looks better in photos than it feels in person. The friend test catches all of that. It's the floor.

If a venue passes the friend test, then we look at three things.

1. Does it do its category well?

We don't ask whether a venue is good at everything. We ask whether it's good at the thing it's trying to be.

A small natural wine bar should be evaluated on whether it's a great natural wine bar — not whether it has the most innovative food in the city. A neighborhood Sichuan restaurant should be judged on whether it serves great Sichuan food, not on whether the room is beautiful. A late-night cocktail spot should be judged on whether the bar program is real, not on whether they also serve dinner.

This sounds obvious. It isn't. Most review platforms penalize venues for not being everything to everyone. We don't. A venue that does one thing exceptionally is more valuable to a user than a venue that does five things decently.

If we can't write a clear, confident sentence finishing "this is the best place in Shanghai for ___," the venue usually isn't in.

2. Is it operationally consistent?

A great meal you can't reproduce isn't a great restaurant. It's luck.

We've turned down venues that had a stunning opening month and then stopped delivering. We've turned down venues that were excellent on a Tuesday and chaotic on a Saturday. We've turned down venues where the chef was inspired and the team behind them couldn't execute it twice.

Consistency is the second filter. We watch a venue across multiple visits, across different days of the week, across different times of day, across different members of our content team. If the experience holds up under that kind of pressure, it earns consideration. If it doesn't, it stays off.

This is where most "hot" venues fall out. A new opening with a great PR push and a packed first month is exciting. It is also, statistically, very likely to be a different place six months in. We wait. The venues that survive the wait are the ones worth recommending.

3. Does it fit our audience?

Unight has 200,000 users with a specific profile. Globally mobile, quality-chasing, suspicious of hype. People who would rather have one great meal than three forgettable ones. Curators in their own lives.

Not every good venue fits that audience. There are restaurants in Shanghai that are genuinely excellent and would be wrong to recommend through Unight — because they're built for a different kind of customer. A high-volume banquet hall doing wedding catering can be operationally brilliant and still not belong on a platform built for individual decisions about taste.

The third filter isn't about quality. It's about fit. We're not the platform for every venue in the city. We're a platform for venues that match how our users actually want to spend their time.

A worked example — the one that didn't make it

Last year a new restaurant opened in a neighborhood I won't name. Beautiful room. Real money spent on the design. The owner had done other successful concepts. The opening was packed. Three different friends asked me within a week if I'd been.

Our team visited twice in the first month. Both visits were genuinely good — the food was thoughtful, service was attentive, the wine list had been chosen by someone with taste. By every short-term signal, it was a contender.

We didn't add it.

The reason was simple: the chef who designed the menu wasn't going to be in the kitchen long-term. We knew it because we knew the industry. The opening team was the A-team. The B-team would arrive in three months. We've seen this pattern many times before. The venue stays beautiful, the prices stay high, and the food quietly drops a level.

Eight months later, that restaurant was a different place. The food was unremarkable. The service had loosened. The friends who had asked me about it in week one didn't ask anymore.

If we'd added it on the opening month, we'd have lied to our users. They'd have walked in expecting the restaurant of the opening week and gotten the restaurant of month nine. That's exactly the kind of broken signal Unight exists to prevent.

So we waited. We watched. And eventually we made the call to leave it out.

This is what curation actually is. Not a list of every place that's nice. A list of every place we'd defend.

What we say no to, every week

To make the standards real, here are the categories of venues that consistently don't make it onto Unight, even when they're competent:

• Venues that peaked and haven't kept up. Reputation can lag reality by years.

• Hype openings where the team behind the launch isn't the team running the kitchen.

• Venues with excellent food and broken service. Service isn't optional — it's half the experience.

• Places that look incredible in photos and feel hollow in person.

• High-volume venues that can't deliver to an individual table.

• Venues that try to be everything. Three menus, four cuisines, one decent dish.

• Anywhere that prioritizes turning tables over the night the customer is having.

None of these are personal. They're operational. A venue can fail one of these filters this year and pass it next year. We re-evaluate. Inclusion is not permanent and exclusion isn't either.

Why this matters

A curation platform that lets in everyone isn't a curation platform. It's a directory with better fonts.

The reason Unight is useful is because the things on it are all worth your time. Not most of them. All of them. That promise only works if we say no, often, to venues that other platforms would be thrilled to list.

The cost of saying no is that we're smaller. We have hundreds of venues, not thousands. We're missing places. Some of them are good. Some of them, we'll add later when they've proven they can sustain it. Some of them, we won't.

The benefit of saying no is the only thing that makes the platform mean anything. When a user opens Unight and asks where to go tonight, the answer they get back is from a list every member of our team has personally stood behind. That's the entire product.

If we ever stop being able to say that, we've stopped being Unight.

Want to be considered for Unight? If you run a venue you believe meets the bar, reach out to our partnerships team. Every submission is reviewed by our content team.