The competition in instant retail is undergoing a significant shift. In the past few years, the focus was on who could open stores faster and cover more ground. But as store networks expand, more practitioners are realizing that without a controllable supply system and stable operational capabilities, scale alone is hard to convert into deterministic growth. For today's sharing, I don't want to dwell too much on macro judgments. Instead, I'd like to stand from the perspective of a frontline practitioner and systematically walk through our real experiences, the pitfalls we've encountered, and the methods we've gradually developed over the years in the instant retail warehouse-store model. Looking back, we didn't have a fully thought-out 'top-level design' from the start; it was more about continuous correction during execution. But in hindsight, certain choices were key to getting us where we are today. From 10 stores to 3,500, In the process of finding certainty amidst uncertainty, we refined the model. At the very beginning, it was simple—we opened only 10 warehouse-stores in a year. Our understanding of the industry was limited then; we just felt it was a direction worth investing in, but we didn't have a mature playbook for execution. However, there was one action we took from day one: while opening front-end warehouse-stores, we simultaneously set up central warehouses. At the time, this seemed like an 'extra step,' but in hindsight, it was the foundation that allowed our entire model to work. If we had only focused on stores, we might have gone down a different path. In the following years, as the industry rapidly developed, our pace accelerated. From the initial dozen or so stores, we gradually expanded to where we are today, roughly at this scale: We now have 22 central warehouses nationwide, achieving broad national coverage; our stores reach 1,800 counties and cities, with a total of nearly 3,500 stores, of which nearly half are self-operated and the rest are partner stores. But after scaling up, we spent a lot of time reflecting on a fundamental question: What exactly are we doing?

Are we a store-opening company?

Or a channel that sells goods?

Or a service provider? We didn't answer these questions all at once; we corrected course through trial and error. It was during this process that we began to realize—if we just stay at the 'store-opening' level, there's no moat. Building a Warehouse Network Making the Business 'Controllable, Replicable, and Scalable' We don't treat warehouse-stores as isolated points but as a 'warehouse network.' That's why we've consistently adhered to the 'central warehouse + front-end warehouse' structure rather than a single front-end warehouse model. In our system, the division of labor is clear. Central warehouses handle centralized procurement, aggregation, and management of heavy and long-tail items, as well as regional inventory turnover. All the 'heavy,' 'slow,' and 'scale-dependent efficiency' parts are concentrated in central warehouses. Front-end stores follow a completely different logic. We want them to be as light as possible, with shallow inventory but fast turnover. They use smaller capital to support higher-frequency sales and fulfillment. Behind this is a simple logic: Bring the complex parts back, and push the simple parts out. It's precisely because we have central warehouses as an intermediate layer that we can achieve higher efficiency at the store level without being dragged down by inventory. But as store numbers grew rapidly, we soon encountered a new problem—if every store relies on human experience for operations, the system can't be replicated. So the second thing we did was to systematically formalize the entire process from 'opening a store to operating it.' We built a comprehensive capability set, from site selection and product selection to daily store operations, using systems for coordination as much as possible. For franchisees, the most direct change is that the barrier to entry is significantly lowered, with integrated management from store opening to operations. Next came the digitalization of supply chain and operational capabilities. We invested heavily in data capabilities. Store replenishment no longer relies on experience but is automatically decided by systems based on historical data and sales trends. When to replenish, how much, and which categories—all are data-driven. At the same time, we integrated the entire operational chain, from product management and order fulfillment to business analysis and financial systems, all running within a single system. We have a consensus internally: Instant retail is essentially e-commerce, and the core of e-commerce is data-driven decision-making. Only when these capabilities are truly built can the business have the foundation for being 'controllable, replicable, and scalable.' When You Have a Warehouse Network, You Truly Enter 'Brand-Efficiency Symbiosis' On this basis, our relationship with brands has also undergone a significant change. Initially, what we did was simple: we sourced products from the supply chain and sold them to users—a typical distribution logic. At that time, we were more of a 'channel.' Later, we began deeper collaborations with brands, such as new product testing, data feedback, and product selection optimization, gradually entering a 'deep cooperation stage.' But after our store count reached 1,500, then over 3,000, covering 1,800 counties and cities, we realized more profoundly that we are no longer a single-point channel but a nationwide instant retail warehouse network. More importantly, this network has a key characteristic—all product operations, pricing rhythms, and marketing actions are centrally controlled by headquarters. What does this mean? It means that once a brand enters this network, it can achieve several critical capabilities: First, rapid coverage at scale. Once cooperation begins, it can cover thousands of stores in a short time, rather than advancing store by store. Second, controllable rhythm. Pricing, promotions, and new product launch rhythms can be executed uniformly, avoiding channel conflicts. Third, complete data recovery. All sales, conversion, and fulfillment data can be accumulated and used to guide decisions in reverse. In other words, we have developed the ability to provide 'certainty' to brands. Meanwhile, we've also undergone a transformation at the operational level. We no longer focus only on 'what products to sell' but increasingly think about 'in what scenarios users consume.' We look at it from three dimensions: One is time, such as 24 hours a day, different times of the week, and different seasons;

One is space, such as different cities and regions;

One is specific consumption scenarios, such as breakfast, travel, and late-night snacks. When we restructure the product assortment this way, the logic changes. Users no longer just buy a single item; they complete a whole set of consumption within a scenario. For example, they might originally buy just a drink, but in a 'late-night snack scenario,' they might form a bundled purchase. This not only increases average order value but also improves overall conversion efficiency. Summary Looking back, our feeling is that instant retail is undergoing a very obvious change. In the early days, everyone competed on 'who opened more stores,' but once scale is achieved, the real differences emerge on other levels: Do you have the ability to make supply more controllable?

Do you have the ability to achieve higher efficiency?

Do you have the ability to form deeper synergies with brands? In other words, this is no longer a simple store-opening business but a competition of systematic capabilities. Looking ahead to 2026, we will continue to focus on several things: First, moderately densify the warehouse network to improve efficiency on top of existing coverage; Second, continue optimizing the 'central warehouse + front-end warehouse' model to make supply smarter and more controllable; Third, further deepen collaboration with brands, not just 'selling goods,' but exploring more systematic growth methods centered on users and scenarios. In this process, we will also more proactively cooperate with more brands interested in instant retail, moving from single-channel cooperation to deeper brand marketing integration—not only helping brands achieve sales conversion but also coordinating efforts in content, scenarios, and channels to drive the integrated output of 'brand, efficiency, and sales.' At the same time, this year we officially opened our To B supply chain system. With the financial and tax compliance system and supply chain capabilities we've accumulated over 4-5 years, we aim to output more to the industry, helping more mid-to-late-stage instant retail merchants achieve more standardized and cost-effective comprehensive supply, and jointly progress with the industry. We always believe that this industry is still in its very early stages, but one thing is becoming increasingly clear—the real competition is not about how many stores you have, but whether you can stably run this system and replicate it. We also hope to work with more brands, supply chain partners, and industry peers to make this endeavor more solid, so that more people can enjoy the beauty of instant delivery to home.