This is an English adaptation of a FoodBud historical article originally published on November 7, 2022.
This article is adapted from analysis produced by GrowthBox Research Group, led by Yanju Xiaosu with researcher Binchao. It examines how Luckin Coffee moved from aggressive, subsidy-led expansion and a 2020 accounting scandal to renewed growth and profitability by reframing coffee as a high-convenience, mass-market beverage business.
Luckin’s core bet was that China had room for a mass coffee market between instant coffee below RMB10 and premium cafe coffee above RMB25.
Before 2017, freshly made chain coffee below RMB20 was limited mainly to McCafe and K COFFEE inside McDonald’s and KFC stores. Starbucks and similar cafe chains had trained consumers to associate coffee with a premium lifestyle, third-place space, service and social status. In October 2016 data, a Starbucks coffee in Zurich cost about 0.1% of monthly disposable income, while in Beijing it was 0.76%, showing a large premium relative to income.
Luckin entered in 2017 with low prices and fast pickup. Its small-store model removed most cafe-space costs and treated stores like flexible production nodes: automated machines, formula-based recipes and limited customer interaction. In effect, Luckin sold freshly ground coffee with the logic of bottled beverages.
That let the brand serve lower-frequency coffee consumers who wanted taste accessibility, value and portability. As scale increased, product quality improved and some consumers from premium coffee networks accepted slightly lower product expectations in exchange for convenience.
Before Luckin, Lu Zhengyao founded CAR Inc. in 2007, listed it in Hong Kong in 2014, launched UCAR in 2015 and merged it into UCAR Group in 2016, which listed on China’s NEEQ with a market value near RMB41.8 billion.
Lu placed unusual emphasis on data. CAR launched its own app in 2011, enabling 24-hour pickup and return. At Luckin, he repeated this approach more aggressively. Before the first store opened, nearly 40 developers from the CAR team built the Luckin app, which became the center of transactions and user experience.
Luckin also recruited heavily from digital, consumer and operations backgrounds: Guo Jinyi and Yang Fei from the CAR ecosystem for technology and growth, Zhou Weiming from Coca-Cola for product and supply chain, Cao Wenbao from McDonald’s for store operations, and He Gang from JD Cloud for algorithms and AI.
Most of Luckin’s digital tools were self-built rather than bought, allowing tight integration with operating needs. Its product and technology team reportedly reached more than 800 people before Lu left in 2020, then stabilized around 250. For comparison, Haidilao’s digital team was around 100 in 2018 and above 400 later; Nayuki and Heytea were around 50-70 in 2018 and above 300 later.
Luckin opened its first store in Beijing Galaxy SOHO in 2017 and completed the journey from founding to IPO in 17 months.
In February 2020, Muddy Waters released a short report alleging hundreds of millions of dollars in fabricated sales. Luckin’s share price fell 80%, nearly US$5 billion of market value was erased, and the company delisted from Nasdaq in July 2020. Lu Zhengyao and parts of the management team left after internal accountability actions.
The report also criticized Luckin’s business model. It argued that the company relied too heavily on discounts, had low loyalty, faced risk from raising effective prices, and needed both higher unit volume and higher realized prices to reach sustainable store economics. It estimated that a store needed to sell 200 items per day and raise average effective price by 43% to become profitable within a three-year lease, with only about 3% store margin.
By 2022 Q2, Luckin reported total net revenue of RMB3.299 billion, up 72.4% year on year. GAAP operating profit reached RMB242 million, compared with RMB16.1 million in Q1.
Monthly transacting customers rose from 7.73 million in 2019 to more than 20 million in 2022 Q2. Store count resumed growth after contraction in 2020. Luckin lost RMB3.212 billion in GAAP net margin terms in 2019, narrowed losses from 2021 and had become profitable in 2022.
The turnaround rested on three operating upgrades:
Before delisting, Luckin had a reputation among many consumers for poor taste. Its menu relied on basic Americanos and lattes, with few standout items other than the Meteorite Latte.
From early 2020, Luckin reworked its product innovation process. It launched Thick Milk, Meteorite, Coconut Latte, Coconut Cloud and Bayberry Exfreezo products. In 2021 alone, it launched 113 new products.
The Coconut Latte sold more than 100 million cups after launch, with monthly sales above 10 million cups. The Coconut Cloud series also targeted 100 million cups.
GrowthBox describes Luckin’s product mechanism in four stages:
About 70%-80% of new-product planning came from market and consumer data, while 20%-30% came from creative ideas and raw materials.
The product management center used public data scraping, competitor research, trend research and Luckin app consumption analysis. The R&D center also scanned supplier materials and proposed ideas based on new ingredients.
In early 2019, Luckin’s product team had only 18-20 people, with 4-5 focused on new products. After the Little Deer Tea project, tea-drink R&D talent remained and brought new thinking to coffee development. The later product team was around 50-60 people, with 2-3 new-product groups of 5-6 people each.
Luckin used a “horse race” structure, with multiple teams independently developing products. Average development time was about two months, but one group could work on 5-6 products at once.
Each R&D group produced roughly 5-6 samples per week and selected 2-3. Overall, 4-6 products entered internal review weekly.
Testing involved 50-60 internal employees scoring taste dimensions such as sweetness, acidity, bitterness and aroma, plus mouthfeel dimensions such as richness, body, viscosity, integration, balance and smoothness.
The Coconut Latte was developed in early 2020 but delayed because its ingredients were unusual for the market. It went through more than seven rounds of internal and external testing over about a year before launching in April 2021.
After testing, Luckin reviewed production efficiency, ingredient cost, profit and operational burden, then created standardized SOPs for stores. New products were typically promoted heavily for one week. If sales reached 50%-60% of a normal latte, the product could stay on the menu; otherwise it was removed after initial materials were used. If it entered the top three daily menu items, support increased and it could become permanent.
Luckin’s product philosophy was summarized by Zhou Weiming: the advantage was not one good product, but a mechanism for quickly producing good products.
Luckin’s store operating net margin moved from -12.5% in 2020 to 20% in 2021 and 30.6% in 2022 Q2.
Its simplified profit formula was revenue minus raw materials, rent and operating costs.
Raw materials were mainly coffee beans, milk and coconut milk. Scale purchasing lowered cost. Luckin’s first self-built coffee roasting plant was expected to reduce coffee-bean costs by another 5%-8%. Estimated ingredient cost was RMB2-5 per cup, excluding delivery, R&D and marketing.
More than 97% of Luckin stores were 20-60 square meter pickup stores, usually on office-building ground floors or mall corners. Average national rent was RMB15,000-19,000 per month, typically around 10% of revenue. Starbucks and Nayuki-style large stores often had rent above 15% of revenue.
Luckin’s early customer acquisition was criticized as buying volume with subsidies, but the company viewed marginal acquisition cost as only the ingredient cost of one cup.
The first app version in October 2017 supported ordering and payment. A new user could download the app and receive a free coffee, costing Luckin roughly RMB2-5 in variable cost. After that, users could invite friends for a 1.8-discount coupon or buy discounted coupon bundles. By a second purchase at RMB5-6, the payment covered ingredient cost; by a third purchase, acquisition ROI could exceed 1.
From 2018 to 2021, Luckin spent RMB3.21 billion on marketing, including advertising, free-product promotion, delivery-platform subsidies and other items, and acquired 92 million transacting users. That implies about RMB35 per app-downloading, registered and paying user. Pinduoduo, expanding through social referrals in the same era, spent RMB77 per app download and registration in 2018 and more than RMB200 in 2020.
By the article’s data, more than 70% of Luckin’s business came through the app and another 20% through mini-program orders. Starbucks’ own online channels at 60%-70% and Heytea’s 60%-65% were seen as high benchmarks.
Luckin also used WeCom after becoming an early test user of WeCom 3.0 in early 2020. It connected WeCom with its self-built SCRM, CDP, BI dashboards and third-party tools. As of April 2022, only 20% of community users entered groups from stores. Luckin had more than 30 million WeChat service-account subscribers, more than 20 million WeCom users and around 16 million community users. More than 80% of community users came from WeChat ecosystem referrals and ads.
Yang Fei said automation in Luckin’s WeChat private-domain marketing exceeded 90%. Daily conversion after coupon distribution in communities reached 30%. Coupons could be distributed based on RFM models, consumption tags, time, discount depth and product category.
Luckin also reduced discounts over time: from a lowest 1.8 discount in 2019 to 3.8 in 2020, 4.6 in 2021 and 4.8-5.6 in 2022. Actual user payment per cup had risen above RMB15.
Luckin’s smart store system automated processes and scheduling.
New employees could often become independently productive after 2-3 days of in-store training, compared with one month or more at Starbucks and Manner. Stores mainly used fully automated SCHAERER and FRANKE coffee machines, with semi-automatic machines as support. Employees followed machine prompts and SOPs through Luckin University and the workstation app.
The smart store system also scheduled full-time and part-time workers each Friday based on order forecasts and peak-trough patterns. Store managers and grid operations staff could redeploy employees through the app when needed.
In-store smart screens calculated production order based on equipment status and incoming orders, then split production, packing, pickup and delivery timing with countdown alerts. Staff did not need to handle cashiering, ticketing or delivery dispatch.
Luckin required a fastest service time of 1.5 minutes per cup. GrowthBox fieldwork found average actual production time below 50 seconds, with theoretical speed up to 100 cups per person per hour per machine.
This automation enabled heavy part-time staffing. In many cities, part-time staff were about 50% of store labor; in first-tier cities, close to 70%. Pay was RMB25-29 per hour, and labor cost was below 15% of revenue. Starbucks and Manner also used part-time staff, but usually within a 20%-50% range.
In 2022, core Luckin customers in first- and second-tier cities bought 3.7-5 cups per week, up from 2.7 in 2021.
Using 2022 Q2 data, assuming 30 operating days per month and RMB15 per cup, Luckin’s directly operated stores averaged 351 cups per day nationally before adjusting for pandemic closures. Luckin had average daily closures of 950 stores in April, 900 in May and 152 in June 2022; adjusting for closures, daily direct-store volume likely exceeded 400 cups.
Competition still mattered. In Shanghai and other first-tier markets, when Manner, Tims and similar-priced brands opened nearby, Luckin stores reportedly lost around 100 cups per day for each added competing store.
Before 2021, Luckin mainly relied on expansion-team resources and experience. Teams had incentives and elimination rules: failure to open a new store for two consecutive months could lead to elimination, and new-store bonuses had four tiers of RMB3,000, RMB5,000, RMB8,000 and RMB10,000. Further rewards and penalties depended on average daily orders and 3-, 6- and 12-month survival rates.
In January 2021, Luckin launched a self-built smart site-selection system. It combined internal data with external service-provider data such as Tencent audience data, AutoNavi maps and delivery platforms. The system mapped foot traffic, consumer profiles, behavior and store locations into a visual “smart map.”
Expansion staff could view scenario analysis such as traffic and opening plans, plus algorithmic forecasts for sales, costs, breakeven, payback and nearby Luckin store cycles.
The model was trained on transaction, geographic, site and traffic data from 7,250 store openings and closures between 2017 and 2020. Luckin also used low-cost delivery stores in new cities to build order heatmaps and identify future sites. By 2020 Q1, after sufficient data accumulation in first-tier and new first-tier cities, these delivery stores gradually converted toward pickup stores.
New stores selected with the smart system had about 20% higher sales than contemporaneous stores opened without it. New-store 12-month survival rose from 83% in 2019 to 95% in 2021, and the 2022 six-month survival curve suggested further improvement.
During the April-May 2022 Shanghai pandemic period, many vacant sites became available. Luckin temporarily added 100 direct stores in Shanghai, supported by its traffic forecasting.
Around the same time as smart site selection, Luckin reopened franchising under a partnership model.
By the article’s data, Luckin had 100% coverage in first-tier, new first-tier and second-tier cities, 99% in third-tier cities and 90% in fourth-tier cities. Further growth required lower-tier markets.
In 2022 Q2, Luckin had 4,968 direct stores in 59 cities, mainly first- and second-tier markets, and 2,227 partnership stores in 231 cities, mainly third- to fifth-tier markets.
In 2022, partnership openings accelerated. By August, nearly 55% of more than 1,000 new stores were partnership stores. The article expected direct and partnership stores among new openings to reach a 3:7 ratio in 2024.
For operators in third- to fifth-tier cities, Luckin’s appeal was that tea competition was intense while coffee remained relatively open. Office-oriented Luckin stores also had less seasonality than tea drinks. The partnership-store payback period had shortened from an average 24 months in 2019 to a national average 14-15 months, with stronger stores recovering investment in 9 months or less.
Luckin required partners to submit sites for headquarters review. Luckin staff assessed traffic, surrounding demand and annual profit potential, then provided analysis.
Lower-tier partnership stores had lower rent and labor costs. Although they paid higher material costs and profit sharing than direct stores, operating net margin could exceed that of higher-tier stores. First- and second-tier stores were usually around 28% operating net profit, while lower-tier stores commonly exceeded 30%.
At the time, the smart site-selection system was mainly used for direct stores. Partnership stores received templated analysis reports and still relied more on partner and expansion-team experience in lower-tier cities. The article expected the system to support partnership stores within a year, potentially improving net margins further.
Luckin sits between two business logics. Starbucks represents the cafe model, built around coffee quality and social-cultural space. Nestle represents the packaged-consumer-goods model, built around availability. Luckin kept some freshly made coffee quality while pushing convenience much further.
That position creates risk. Compared with traditional cafes, Luckin’s small stores lack offline space, making it harder to build deep customer relationships or raise customer value beyond drinks. Attempts such as Luckin Trend Products, Ruijigou and Ruihuasuan had limited results, and non-beverage categories contributed only about 5% of sales. By contrast, Starbucks’ non-beverage business contributed more than 30% of total revenue, with retail products above 20% of profit.
Compared with packaged beverage giants, Luckin’s store network remained small. Even with nearly 10,000 stores nationwide, it was far below FMCG penetration. Lower-tier markets were also more complex and relationship-driven, and data plus process alone might not solve distributor and local management challenges. Luckin had already faced copycat issues in lower-tier markets, affecting brand strength and causing a temporary pause in new partnership openings.
The biggest operating risk was product concentration. In a given month, Luckin’s top three products could contribute 40%-70% of store sales, and Coconut Latte alone could contribute 30%-50%. Competitor stores usually had classic Americanos and lattes among top sellers, with no single product above about 15% of sales.
Luckin was still highly dependent on hero products. Industry operators believed a beverage product life cycle rarely exceeds three years. In 2021, only one in ten Luckin listed products remained on the regular menu, while more cautious tea and coffee brands could reach menu retention above 50%.
The conclusion: Luckin had proven that China’s mass, convenient, freshly made coffee market was real, and it had rebuilt a stronger operating model after delisting. But becoming a long-term category leader would require continued self-disruption, especially in product consistency and lower-tier-market execution.
Note: IPO, delisting, financial, margin, payback and forward-looking store-mix figures are historical figures from the 2022 source article.