Lite-version red packets aren't growth, they're a harvester for the algorithm and monetization
Once the demographic dividend was gone, the point of red packets, Lite apps and coupon drops was never growth — it was using monetization and the recommendation algorithm to absorb the traffic. In the AI era paid acquisition is close to useless, and growth goes back to model capability and mindshare marketing.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Growth hacking expired in 2026 because the cheap traffic is gone
Tom argues growth hacking is out of date as of 2026. The global mobile internet has been fully cultivated; the giants have it locked up, with ByteDance and Tencent leading domestically and Meta and Google holding seventy to eighty percent of traffic overseas. Monetization is so thorough that organic traffic has dried up, while platform risk controls and users' own instincts keep getting sharper, so the high-leverage freebie-farming tricks no longer work. In the AI era product supply is not scarce, which means growth matters more than ever — but the old methods will not deliver it.
— TomCoin rewards poison the recommendation signal in a text feed
Toutiao (头条) copied Qutoutiao (趣头条) and built a Lite version, then discovered that text-and-image news is a bad fit for the earn-money model: users read content to complete a task and collect coins, not as a genuine expression of interest. They pick the short pieces and swipe away fast, which badly corrupts the recommendation algorithm. Free web novels and short video are the opposite fit, because narrative content tolerates inserted ads, and short video has both longer sessions and higher monetization efficiency. That is why Midu (米读), Fanqie (番茄) and the Douyin and Kuaishou Lite apps worked.
— TomLong-video giants' Lite apps were doomed for three structural reasons
The Lite versions built by iQIYI, Tencent Video and others all failed, and Tom breaks the failure into three layers. First, the organization does not fit: long-video companies run on a chief-editor and producer system, with no recommendation algorithm and no short-video content production pipeline. Second, content supply is thin — clips from films and variety shows alone cannot fill the hours an up-and-down swipe feed demands. Third, the business models conflict: free ad-supported viewing eats into paid membership. The Douyin and Kuaishou Lite apps succeeded because their content covers everything and their monetization absorbs the traffic — ads, livestreaming, e-commerce.
— TomVirality is math: above a K-factor of 1 it can take off
Virality can be quantified. The K-factor has to exceed 1 before volume is even possible, and above 2 you have a very good product. Travel Frog (旅行青蛙) was measured at a viral coefficient above 10, which is exponential growth; Toutiao's Baiwan Yingxiong (百万英雄) was around 3. Years ago Tom worked on a Taiwan product that exploited a Facebook login loophole where one user could spawn seven, and Facebook itself held it up as a case study. But opportunities like these depend on the WeChat or Facebook social graph, or on timing you cannot plan for. Because ByteDance has no social graph of its own, it has struggled ever since to manufacture another viral hit.
— TomWith no social graph, ByteDance went after phone-level permissions instead
After WeChat blocked it, ByteDance's products turned sharing into an exploitation of the phone's underlying permissions: for a few mao, users are induced to open up their contacts, photo library, Bluetooth and more. The Douyin contacts that appear when you tap share come from the address book and from face recognition on the photo library; Qishui (汽水) can pop up the moment you put your earphones in, using RT wake-up technology to detect Bluetooth state. The photo library, which carries GPS locations and relationships between people, becomes a data source. Tom's warning to users: be careful about tapping ‘Always Allow’.
— TomRedemption rate, not coupon volume, is how Douyin overtook Meituan
Douyin's local services business made coupon redemption rate its north star metric. After lifting it from under twenty percent to a considerably high level, on several days late last year its Beijing in-store order volume exceeded Meituan's in-store business. A higher redemption rate means the user's mindset has shifted from ‘I happened to scroll past a deal’ to ‘I am going there to spend’, and that is where the standalone app Doushengsheng (抖省省) came from. For merchants and for users, Douyin's coupons now sit ahead of Meituan's by default — especially in lower-tier markets, where for many people their first group-buy purchase happens on Douyin.
— TomThe red-packet middle platform is really the app family promoting itself
ByteDance built its red-packet middle platform in 2019, packaging user acquisition, engagement and traffic routing into an SDK that every product in the family could plug into. Among the dozens of tasks running over Spring Festival, tasks that routed traffic to Dongchedi (懂车帝), Xingfuli (幸福里) and Haitun Gupiao (海豚股票) got heavier coin weighting, because their Apple users had not been washed over yet, so new users were plentiful and the apps could be lifted to the top of the App Store chart instantly. Reaching No. 1 on China's Apple chart takes roughly 300,000 new daily installs; using red packets as leverage to move volume from one hand to the other, ByteDance took six or more of the top ten.
— TomBuying users is a false proposition for AI products right now
Tom argues that at this stage, buying traffic for an AI product is a false proposition: model capability determines what the user perceives, and as long as the leader is 5% ahead, users will only pay for the best. Doubao's overseas version bought 90 million users over the past 3 months, but its DAU is only 16 million — retention is poor because the model cannot match its American competitors. So real growth is integrated marketing and mindshare: the founder's personal brand, podcasts, Twitter, Product Hunt — getting users to spread the word themselves because ‘you are the strongest’.
— TomIn their own words · checked verbatim
Taobao Deals (淘宝特价版) also sells goods — it solves exactly the same core problem the main Taobao app does.
淘宝特价版，它也是卖货啊，它跟淘宝主端解决的核心问题是一样的。
Tom1:02:36
Virality has a statistical concept called K-factor, the K-factor effect: when one of your users can bring in more than one user, that tells you virality can take off.
裂变就是它有一个统计悬概念叫Kfactor，叫K因子效应，就是当你一个用户能带来大于一个用户的时候，说明裂变就能起来。
Tom1:23:21
The WeChat red packet really is a hell of an invention. Honestly, the most badass viral loop in history really is the WeChat red packet.
微信红包真的是个很牛逼的发明。 真的你说历史上最牛逼的裂变，就真的就是微信红包。
Tom1:26:35
One thing I think ByteDance's products do in an extremely over-the-top, extremely savage way is sharing.
我觉得自节系产品做的非常过分，也非常野蛮的一个点，就是分享。
Tom1:35:02
Buying traffic for an AI product is actually a false proposition at this stage — your product's own strength comes first.
AI产品买量其实在现阶段来说是个伪命题，你产品利益是第一的啊
Tom2:52:57
Figures
| Active Apple devices in China | roughly 200 million-plus | 2:27:34 |
| Daily new installs needed for No. 1 on China's overall Apple chart | about 300,000 | 2:27:34 |
| Travel Frog's viral coefficient | greater than 10 | 1:23:21 |
| Users spawned per user via the Facebook loophole on the Taiwan product | 1 user spawned 7 | 1:24:21 |
Glossary
- K-factor (viral coefficient)
- How many new users one user brings in on average; above 1 and self-driven growth holds.
- RY1 / RY2 / RY3 (acquisition ROI, gross-margin ROI, net-profit ROI)
- RY1 is customer acquisition cost against LTV; RY2 adds variable costs; RY3 adds fixed costs on top.
- Red-packet middle platform
- ByteDance's general-purpose red-packet system, packaged as an SDK so every product can route traffic and drive engagement.
How to listen
People working in growth, product and monetization, plus founders and investors who want to understand the Douyin-Meituan coupon war and the logic of AI customer acquisition.
The middle section on the failed iQIYI and Tencent Video Lite apps can be fast-forwarded.