Philip Chen is the CEO of GMA (Gentlemen Marketing Agency). Since 2013 he has run China launches for fitness apps, connected equipment brands and wearable makers, and he tracks Keep’s quarterly filings the way other people track football scores.
Three years ago I wrote about digital fitness in China. Almost every number in that piece is now wrong. Keep went public, lost users, cut its own revenue on purpose, and finally made a profit. Smart mirrors went from hyped gadget to quiet niche. Gyms opened with no staff at all, just a QR code on the door. And the biggest fitness audience in the country turned out not to be in any fitness app.
This piece covers the digital side only: apps, content, connected hardware, AI coaching, wearables. For the clubs, the membership card economics and the closures, read the wider fitness market in China first.
The Numbers, And Why The Old Forecast Missed
For years everyone quoted the same forecast: China’s online fitness market was worth about 370.1 billion yuan in 2021 and would reach 895.8 billion yuan by 2026, with online fitness penetration hitting 57.4%. That number was written during the lockdown years. It assumed people who started training at home would keep doing it.
Most of them did not. The real picture is smaller, and it splits in two.
QuestMobile’s sports and health report puts every fitness and sports app in China at about 67.78 million monthly active users combined. Keep leads with roughly 15.09 million, then 天天跳绳 (Tiantian Tiaosheng, “jump rope every day”) at 4.9 million, 糖豆 (Tangdou, a dance fitness app popular with people over 50) at 3 million, and 咕咚 (Codoon, a running tracker) at 2.87 million. That is the whole app market. It is not big.
The other number from the same report: about 491 million people actively follow fitness influencers. Seven times more people watch fitness content than open a fitness app.
They are also not the same people, and this is what most foreign brands get wrong. App users skew young and urban: 43.1% under 30, 52.7% in tier-1 and tier-2 cities, and 47.2% spending over 2,000 yuan a month online. The content audience skews the other way: 30% over 46, 49.9% in tier-3 cities and below. The app is where you find the 28-year-old in Shanghai. The video feed is where you find her mother in Hefei. They buy different things.
Keep: The Year It Finally Made Money
Keep is still the app people name first, and it earned that. It offers workout plans, live classes, running and cycling tracking, and a feed where users post their progress. Its Chinese slogan is 自律给我自由 (zilv gei wo ziyou), roughly “discipline gives me freedom”. It listed in Hong Kong in July 2023.
The years after the IPO were rough, then interesting. Keep published its 2025 results in March 2026 and the headline was its first annual profit ever: adjusted net profit of 25.22 million yuan. Revenue was 1.637 billion yuan, down 20.7% from the year before. Gross margin climbed 5.5 points to 52.2%. Average monthly active users fell to 21.8 million, down 8.15 million in a single year.
Read that again. Revenue down a fifth, users down sharply, and the company made money for the first time. Keep did not stumble into this. It killed low-margin business on purpose, mostly cheap hardware and discounted content, and kept the paying core.
The revenue split tells you what Keep actually is now:
- Own-brand fitness products (treadmills, bikes, bands, smart scales): 778 million yuan
- Online membership and paid content: 680 million yuan
- Advertising and other: 179 million yuan
Keep sells more equipment than subscriptions. It is a hardware company with an app attached, not the other way round. Its most valuable asset for you is not ad inventory, it is a user base that has already proved it will pay for fitness.
Where The Coaches Went: Douyin, Bilibili and Xiaohongshu
Ask a Chinese personal trainer under 35 where he works and you often get an odd answer: the gym floor, three days a week, and a phone camera the rest of the time.
The reason is arithmetic. A gym trainer sells session packages and keeps a slice after the club takes its cut. He is capped by hours in the day and by how many members walk past him. Online he is not capped. One hour of filming sells four ways at once: platform traffic revenue, affiliate commission on the gear he uses, a remote coaching plan sold to people in cities he will never visit, and a 私域 (siyu, “private domain”) WeChat group charging a monthly fee for check-ins and meal feedback. A trainer with 80,000 followers and a tight WeChat group out-earns most club trainers in his city, and he keeps the client if he changes gym.
Each platform does a different job, and brands should budget for them differently.
Xiaohongshu is the research layer. Fitness and sports search on the platform grew about 15% year over year, and the audience is narrow and rich: roughly 77% women, mostly 26 to 30, first and second tier cities, over 95% with medium to high spending power. They will pay for private coaching and boutique classes. They also spot a sales pitch instantly. What works is honest before-and-after posts, clear form tutorials, and calm video. What fails is a glossy brand film.
Douyin is the selling layer. Mid-tier influencers, not the biggest names, drive most of the actual sales, and they do it in livestream. Douyin’s recommendation feed also pushes fitness content to people who never searched for it, which is how a brand reaches that older tier-3 audience the apps never touch.
Bilibili is the one most Western brands skip, and for equipment that is a mistake. Bilibili users watch long. A 40-minute 跟练 (genlian, “follow along and train”) video runs while the viewer trains in the living room, comments scrolling across the screen. If you sell a bike, a rower or adjustable dumbbells, one creator using it for 40 unedited minutes does more than ten polished 15-second clips. Flaws show up in a long video, so only send your good product there.
A Client Story: The 380,000 Yuan Lesson
Noora is Finnish. She runs the China business for a Nordic brand selling a connected indoor bike and a chest heart-rate strap, both tied to the company’s own app. She came to us having spent about 380,000 yuan on paid app installs over four months. Cost per install: 28 yuan. Seven-day retention: 6%. She was filling a bucket with a hole in it.
The mistake was not the budget. It was the order. She was asking Chinese users to download a foreign app before they had any reason to trust the hardware. Nobody in China downloads an app to evaluate a purchase. They read Xiaohongshu, they watch someone use the thing, then they buy on Tmall.
We flipped it and killed the install campaign. The money went into twelve mid-tier Xiaohongshu creators posting real six-week training logs, three Bilibili creators doing full-length follow-along rides, and two Douyin livestreams at the end. We built a WeChat mini-program for warranty registration, class booking and a buyer community, and we made the heart-rate strap sync into Huawei Health and Keep instead of demanding the brand’s own app.
Six weeks: just over 1,400 units sold across Tmall Global and Douyin, the brand’s first real sales month in China. The mini-program picked up 3,200 registrations, and 41% of buyers opened it at least four times in the first month. Install budget: zero. It worked because Chinese users never had to leave the tools they already use. The app was the last step, not the first.
Connected Gyms: Nobody At The Front Desk
The most visible piece of digital fitness in China is not an app. It is a room.
无人健身房 (wuren jianshenfang, “unmanned gym”) are small 24-hour rooms with no staff. You buy a card in a WeChat mini-program, scan a QR code at the door, the electronic lock opens, and the system knows who is inside. Some use facial recognition instead. Booking, payment, door control and the monthly bill all live in the mini-program, and the operator watches it from a phone.
The model spreads fast because it is cheap. No front desk, no sales team, no shift rota. Jinan alone now has more than 70 chain-run unmanned venues, plus a long tail of independent shops. 铁兽运动 (Tieshou Yundong, “Iron Beast Sports”) reports over 600 franchised locations across more than 100 cities and 1.17 million registered users. These rooms sell by the session or by the month, often a few yuan for an off-peak hour, never an annual card.
There is a catch, and the Chinese press is on it. Unmanned means unsupervised. Cramped rooms, patchy camera coverage, equipment nobody inspects, no first aid kit, no AED, nobody to help if someone gets hurt at 2am, and reports of faulty locks trapping users inside. Regulation is coming. If you sell equipment, insurance, sensors or access hardware into this segment, safety and remote monitoring is the pitch that lands in 2026. For the club and studio side, see our pieces on boutique fitness studios in China and how big the fitness industry in China really is.
Home Equipment After The Bubble
The 智能健身镜 (zhineng jianshen jing, “smart fitness mirror”) shows exactly what happened to home fitness in China. A screen that looks like a mirror, with a camera that watches your form. FITURE launched the category here in 2020 and everyone piled in: Yijian, Leke, Codoon, with Huawei, Xiaomi and Baidu circling.
Then the confinement years ended. FITURE cut more than 500 staff in July 2022, most of them front-line China roles, and four VPs left. Mirrors sit between 4,000 and over 10,000 yuan, and the pattern has held since: the higher the price, the fewer the units. Industry white papers still project millions of mirrors sold and growth above 40% a year. Treat those as sales targets written by people selling mirrors.
What survived the bubble is less glamorous and sells better:
- Smart jump ropes, scales and resistance bands under 500 yuan, bought on impulse from a short video
- Folding treadmills and walking pads that fit under a sofa in a 60 square metre apartment
- Adjustable dumbbells, which solved storage rather than adding software
- Connected bikes, the one big-ticket item that held on, because riders will follow a class
The lesson for anyone selling hardware into China: apartment size beats feature list. A product that folds, stacks or hides wins in a market where the average urban living room is small.
AI Coaching In 2026: What It Does And What It Does Not
Every Chinese fitness app now says it has AI. Most of it is a plan generator with a chat window. Keep went further. In April 2026 it launched its own sports model, Keepace.ai, alongside Keep 9.0, built on ten years of training data with input from sports physiology researchers at Capital University of Physical Education.
Here is what it genuinely does. It writes a plan from your body data and from the equipment you actually own, so it does not prescribe a cable machine to someone with two dumbbells. It answers training questions in plain language instead of gym jargon. It reads heart rate, cadence, pace and training load, flags when you are pushing into injury territory given your history and fatigue, and tells you to rest. A coach persona called Kaka reads photos of your meals and folds food, sleep and training into one recommendation.
Here is what it does not do. It cannot watch you lift in real time and stop a bad rep. Form checking still works from a still image or a short clip after the fact. It cannot feel whether the right muscle fired, which is most of what a good coach is for. Complex lifts like the snatch and the clean and jerk are beyond it. And it will not chase you when you skip three weeks, which is the actual reason people pay a human.
So AI in Chinese fitness replaces the beginner programme and the nutrition spreadsheet. It does not replace the coach. For a brand, AI is now the retention layer: it is what keeps someone opening your app in month four, and retention decides whether your China business is a one-off hardware sale or a subscription.
The Wrist: Huawei and Xiaomi Own It
Wearables tie all of this together, and this is the one market where the numbers went up hard. IDC put China’s wrist-worn device shipments at 73.9 million units in 2025, up 20.8%. Smartwatches took 50.61 million of that, up 17.2%. Bands took 23.29 million, up 29.4%. Bands growing faster than watches means a lot of first-time buyers came in at the cheap end, helped by government trade-in subsidies.
The share table is brutal for anyone else:
- Huawei: 25.1 million units, 34% share, up 18.6%
- Xiaomi: 20.4 million units, 27.6% share, up 47.8%, the fastest grower in the top five
- BBK (mostly children’s watches): 12.4%
- Apple: fourth, lifted by subsidy-driven price cuts
- Honor: 2.3%
Huawei and Xiaomi together hold over 61% of Chinese wrists. If you sell a fitness product here, you are not competing with them, you are plugging into them. A strap that syncs to Huawei Health, a bike that reads sleep data from a Xiaomi band and eases off the session, an app that pulls step counts without asking for a second device: that is table stakes. A closed foreign app that refuses to talk to Huawei Health looks broken to a Chinese user.
Running feeds the same loop. Road race demand is far past supply, entry ballots for the big city marathons run at a few percent, and people who miss a slot train anyway, on an app, with a watch. If running or outdoor gear is in your line, see our notes on the outdoor sports boom in China’s cities and on the most popular sports in China, or on strength training in China if your gear sits on the barbell side.
Three Ways To Make Money, And What Each One Really Costs
Every brand I meet asks the same thing: subscription, hardware, or e-commerce commission. Keep’s own accounts answer part of it. Here is how the three compare.
| Model | How it works in China | Real-world reference | Main problem | Best fit |
|---|---|---|---|---|
| Subscription | Paid membership inside an app or mini-program: plans, live classes, video library | Keep membership and paid content: 680m yuan in 2025 | Users cancel fast and free content is everywhere. You pay for retention every month | Brands with real coaching content and a weekly reason to open the app |
| Hardware sales | Sell the product on Tmall, Tmall Global, JD or Douyin. One payment, no recurring revenue | Keep own-brand products: 778m yuan in 2025, its largest line | Margin squeezed at every 618 and Double 11. Repeat purchase is years away | Equipment and wearable brands with a price or design edge |
| E-commerce commission | Creators and KOC sell on affiliate links and in livestream, you pay a cut per sale | Douyin affiliate rates for fitness gear run roughly 10 to 30% by category and creator tier | You do not own the customer. Stop paying and sales stop the same week | Market entry and testing demand before committing to inventory |
Most brands that do well here run two of the three at once: sell the hardware, attach content that gives people a reason to come back, and use commission-based creators to open the door in year one. Keep does exactly this, and it still took a decade to reach a 25 million yuan profit. Plan accordingly.
What This Means For Your Brand
Five practical points, based on what works for our clients right now.
1. Do not buy app installs first. Content, then marketplace, then app. In that order. The client above paid 380,000 yuan to learn it.
2. Build the mini-program before the app. Booking, warranty, community and repeat purchase all live in WeChat here. Our guide to WeChat Mini Shop for foreign brands covers the commerce side.
3. Sync with Huawei and Xiaomi. If your device cannot read the watch already on the user’s wrist, you look incomplete.
4. Budget three creator tiers. Top names for trust, mid-tier for sales, KOC (regular users paid a small amount for honest reviews) for everyday proof. Add Bilibili if your product needs more than 15 seconds to explain.
5. Offer a low-commitment entry point. A trial, a monthly plan, a single-session price. This market has watched too many prepaid chains close to trust an annual card.
Frequently Asked Questions
Do we need a Chinese entity to sell connected fitness equipment or run a fitness app in China?
It depends on the piece. Equipment sells cross-border through Tmall Global or Douyin without a local entity, and that is the cheapest way to test demand. An app is different. Hosting content on Chinese servers, taking payments inside a mini-program or collecting health data all need a local entity and the matching licences. Health data is tightly regulated, so budget legal time. Start cross-border, set up locally once the numbers justify it.
Should we advertise on Keep or build our own presence?
Keep is one channel, not a strategy. It sells advertising, co-branded content and equipment placement, and the users left are people who already pay for fitness. But Keep lost 8 million monthly users in 2025 and shrank its revenue by choice. Use it for credibility with committed trainers, and put the bulk of the budget where the 491 million content viewers are: Douyin, Xiaohongshu and Bilibili.
Is AI coaching worth paying to build into our product for the Chinese market?
Yes, if you treat it as retention rather than a headline feature. Chinese users now expect an app to generate a plan from their data and adjust it, because Keep gives that away. What still gets people to pay is a human in the loop: a coach reviewing the AI plan weekly, or a WeChat group where someone answers. Pure AI coaching with nobody behind it churns fast.
Our device has its own app. Do Chinese users really need it to work with Huawei Health?
Yes. They already own a Huawei or Xiaomi watch, and they will not wear a second device to feed your app. If your data stays locked inside your own software, Xiaohongshu reviewers say so, and that review is the first thing the next buyer reads. Integration is a sales argument here, not an engineering nicety.
We Can Help You Reach Chinese Digital Fitness Users
GMA has spent over ten years launching fitness apps, connected equipment and wearable brands in China, from the first Xiaohongshu post to the Tmall Global store and the mini-program behind it.
We work the digital side: creator campaigns on Douyin, Bilibili and Xiaohongshu, mini-program builds, marketplace operation, and the data integrations Chinese users expect.
If you sell a fitness app, connected hardware or a wearable, get in touch with our team for a free consultation.
Sources
- Keep 2025 annual results, 25 March 2026, Sina Finance
- Keepace.ai and Keep 9.0 launch, 24 April 2026, Sina Finance
- IDC China wrist-worn device shipments, full year 2025, published 10 March 2026, Sina Finance
- QuestMobile, 2025年运动健康行业报告, questmobile.com.cn
- Unmanned 24-hour gyms and their safety gaps, 11 May 2026, Tencent News
- 华经产业研究院, 智能健身器材行业白皮书（2026）, sohu.com
