As Luckin Opens in Grand Central, It's Time to Stop Calling It a Coffee Shop
Mark Sage - 6 min read - 27/08/2026
Howard Yu wrote an excellent article recently on Luckin Coffee—about how a company once expelled from the Nasdaq for a $300 million fraud is now opening stores in Grand Central Terminal, selling $1.99 lattes within sight of Starbucks.
It was a compelling piece, but what stayed with me wasn't the comeback narrative. It was Yu's observation that Luckin's real engine isn't the coffee or the stores—it's the platform underneath them.
That left me with a practical question: if the platform is really that interesting, what does it actually feel like to use?
I happen to be in Hong Kong. So I downloaded the app.
Ignoring the language challenges I had as an English speaker, and the fact the app was pretty much all in Chinese for me, navigating the menu was still surprisingly easy. The process of ordering, personalising, selecting the store and paying was intuitive, including one click payment across to AliPay. The order submitted, I simply waited for Coconut Latte to arrive, and then scanned my QR to let the platform know I'd collected it.
Throughout the process, what struck me more than the coffee (which was itself great) was the realisation that I was not sure I had just visited a coffee shop.
I had interacted primarily with a platform. The physical store was simply the place where an instruction sent through that platform became a cup of coffee. The single staff member operating the store was friendly, but they were executing a ticket the system had already generated, priced, and routed. The store did not take my order. The platform did.
That distinction is the entire story.
Not a Retailer with an App
Most retailers begin with the physical business and digitise pieces of it. Store first, then website, then app, then loyalty, then delivery, then personalisation. The mental model remains the store. Digital is another channel through which customers interact with it.
Luckin was built the other way around.
Identity, ordering, payment, offers, and behavioural data were digital from the beginning. The store was designed as a fulfilment endpoint. Today that endpoint happens to be a person in a 20–60m² space making coffee. But architecturally, it could be (has been / is) a vending machine, a delivery rider, a drone, a robot, or something that does not exist yet. The store is not the system. It is an interface to the system.
Once you see it that way, the operating model makes sense.
The obvious benefit of Luckin’s architecture is transaction efficiency: no traditional till, tiny stores, minimal seating, fast ordering, high throughput. But that is actually the least interesting benefit.
Every transaction also becomes a signal.
Luckin does not merely know how many coconut lattes it sold yesterday. It knows who tried one, what they previously bought, whether they returned, whether they bought it again, what else they bought, where adoption differs by location, and how behaviour evolves over time. When you have pretty much every customer identified, you do not guess cannibalisation from a price promotion. You do not infer whether a product is popular from aggregate sales data alone. You know, within hours, what happened to individual behaviour.
This is not a loyalty programme. Many brands have loyalty programmes today. This is a customer platform, and that changes what the business can do.
Data is the product
The CRM industry has spent years talking about using data to determine what offer to give you. Luckin is demonstrating something more interesting: using customer data to determine what product to make you.
The coconut latte was not a marketing decision. It was a product development decision informed by platform data. Yu notes that Luckin introduced it in 2021 to soften coffee’s bitter taste for customers graduating from milk tea. The platform had already identified the gap in the market by watching how identified customers behaved.
Whereas personalisation asks "what should I sell Mark?", the customer platform asks what should we build because of what we are learning from Mark and millions of customers like him?
The first optimises communication. The second can change the company.
And that data-driven approach doesn't stop at the menu.
The customer platform enables Luckin to connect the whole business digitally through data, and that not only supports product development, but also operational development.
Traditional businesses tend to inherit their operating model and therefore inherit its economics. The business costs X to operate, so the product must be priced at Y.
But for Luckin, their data native approach means they can reverse it. The product needs to cost Y, so the business must be redesigned until X makes that viable.
Tiny stores, digital ordering, automated replenishment, supply-chain integration, and recipe engineering are not independent efficiency initiatives. They are consequences of the customer proposition. As Yu describes, Luckin moved down the supply chain—green bean processing in Yunnan, roasting plants in Kunshan and Qingdao—because controlling cost end-to-end is the only way to make the model work at the price the platform demands.
Luckin does not ask technology to make its existing business model more efficient. It uses technology and data to make a different business model possible.
Digital Extension vs Digital Native
McDonald’s is probably the traditional global retailer best positioned to make this type of transition. It has explicitly pointed to China and its extremely high digital penetration as a model for their other markets. Its global customer platform is becoming increasingly central to how it operates.
But there is a fundamental architectural difference.
McDonald’s started with tens of thousands of restaurants, franchisees, kitchens, drive-thrus, legacy processes, established economics, and decades of operating assumptions. It is layering a customer platform back across an extraordinarily successful physical operating model.
Luckin got to design the physical operating model around the platform.
That difference showed up in a painful way earlier this year. McDonald’s management identified a decision to shift the promotional balance from its customer platform offers to its mass value menu as a “bad trade.” Its best customers - the platform users - reduced frequency.
The organisation learned and corrected. But it took a quarter.
That is not a story of slowness. It is a story of different learning cadences. One company is increasingly building a platform capable of learning from millions of customers. The other was architected around continuous learning from millions of customers.
It is worth noting though that there are different ways to be “good at digital”.
Starbucks, for years the poster child for loyalty and mobile app innovation, largely digitised the customer experience layer. Payments, stars, knowing your order, the warmth of the “third place” extended into an app. The digital layer wraps around a fundamentally physical, experiential business.
Luckin digitised the operating system itself. The app does not wrap around the store; the store executes instructions from the app.
That strategic tension is now visible on the street. Starbucks leans into knowing you by name and giving you a place to sit. Luckin opens next door with tasty drinks at roughly half the cost, in a box that exists primarily to fulfil orders generated elsewhere. Starbucks is a brand encumbered by a legacy it is still monetising. Luckin is a platform that happens to serve coffee.
That's not to say that Starbucks is wrong and Luckin is right, they just target different market segments or needs. Whereas Starbucks has doubled down on its 'third place', Luckin was laser focused on one need: to get a coffee into the customer's hand, fast. Starbucks themselves have acknowledged this difference in positioning, and have recently announced the phase out its mobile order and pickup only outlets; as CEO Brian Niccol put it: “We found this format to be overly transactional and lacking the warmth and human connection that defines our brand.”
So whilst Luckin runs more stores, and sells more drinks by volume within China, Starbucks is still able to maintain a higher price point because of that difference in offering - the warmth and human connection.
There is though one structural caveat. In China, delivery infrastructure is so dense that the store can operate almost purely as a fulfilment node. In markets without that density—whether New York, or elsewhere—the physical store carries more weight. Customers may need to pick up directly, or the store may need to do more of the customer acquisition work that super-app mini-programs handle in China.
That though does not change the architecture, it likely just changes the weighting of the endpoint. The platform is still the front door, but the physical box may need to be slightly more visible than it is in a city like Chengdu or Beijing.
Either way, now that operating model is being exported, the question is not really whether Luckin conquers America. It might or it might not. The more interesting question is what happens when companies built around this architecture begin competing directly with companies trying to retrofit it.
That is much bigger than coffee.
For decades, the physical network was the moat. Stores, locations, supply chain, operational scale, brand. But what took Starbucks almost 30 years to build, Luckin has built - by volume - in less than ten. And their customer platform was key to that success.
This means the moat may be moving— from physical stores, to the digital platform — enabling those organisations like Luckin to more rapidly sense customer behaviour, learn from it, and change what it does.
McDonald’s clearly sees this. That is why China matters so much to them, and why their digital and customer platform investment matters.
But Luckin demonstrates what the destination looks like when you do not have to transform into it.
You were born there.