背景简单说：七年海外用户增长，腾讯、头部游戏大厂、字节跳动、头部游戏出海公司四段，App 与 PC/Web 两端都完整操盘过。这个站不是简历，是工作笔记：投放该怎么判断、AI 怎么长期接管重复劳动、哪些做法被证伪了——能公开的部分都会放在 Learning 里。Background in one line: seven years in global user acquisition across Tencent, a top-tier games company, ByteDance and a leading games publisher, on both App and PC/Web. This site is not a résumé — it's a working notebook: how to judge paid media, how an AI workflow takes over the repetitive parts for the long run, and which practices turned out wrong. Whatever can be shared goes under Learning.
海外 Google、Meta、TikTok、Bing 与国内主流渠道都完整操盘过，App 和 PC/Web 两端都做。方法上，我把归因系统、投放链路、媒体采买当一个整体来做，而不是三件事。PC/Web 端投放是市面上较稀缺的能力——多数投放人才只做过移动端——也正好是 AI 产品的主要战场。I've run Google, Meta, TikTok, Bing and the major China channels end to end, on both App and PC/Web. My approach treats attribution, the conversion funnel and media buying as one system rather than three jobs. PC/Web acquisition is a scarce skill — most UA people are mobile-only — and it's where AI products live.
用户增长（UA，买量）就是付费在 Google、Meta、TikTok 等平台投广告，把目标用户带进来，变成注册和付费。它和做内容、等口碑的区别只有一条：今天花钱，明天就有数据。一旦算清「花 1 块能回来多少」，它就可以按可控的方式加码。User acquisition — paid growth — means paying Google, Meta or TikTok to put your product in front of the right people, and turning them into signups and paying users. Its one real difference from content and word of mouth: spend today, see data tomorrow. Once you know what a dollar in returns, you can scale it in a controlled way.
同一套验证过的打法可以在预算、渠道、市场三个维度同时放大，这是买量区别于其他获客方式的地方。One proven playbook scales along budget, channel and market at once — that is what sets paid growth apart from other acquisition channels.
不达标就放大，等于放大亏损。判断什么时候停，比会加预算重要。Scaling an unproven setup scales the losses. Knowing when to stop matters more than knowing how to spend.
你的产品现在该不该投、投不投得起。账算不过来，我会直接说「先别投」，并说明先修什么。Whether your product should spend now, and can afford to. If the math fails, I'll say "not yet" and tell you what to fix first.
数据追踪和算账体系搭在你自己的账户上。之后不再合作，这套基建仍然是你的。Tracking and reconciliation are built on your own accounts. If we part ways, it stays yours.
小预算验证 ROI 达标后，沿预算、渠道、市场三个维度放大，获客从碰运气变成可以加码的生意。Once a capped test proves ROI, we scale along budget, channel and market — acquisition stops being luck.
付费投广告，老板通常关心三件事：钱花出去有没有赚回来；数据这么多信哪个；养团队或请代理值不值。下面按这个顺序回答。If you pay for ads, three questions matter: is the money coming back; which numbers can be trusted; is a team or agency worth the cost. Answered in that order.
广告数据和你自己后台的真实注册、付费数据接在一起算，不信广告平台自己汇报的数字。这样才能回答「这个渠道到底赚不赚钱」。Ad data is reconciled against your own backend — real signups, real payments — not the platform's self-reported numbers. That is the only way to answer whether a channel is profitable.
每一条广告每天过一遍：哪条在亏、哪条在赚、哪条数据反常。反常的先交叉查证再动手，既不放过真问题，也不被假警报带偏。Every ad is reviewed every day: what's losing, what's earning, what looks off. Anomalies are cross-checked before anything moves.
这些工作过去需要 3–5 人加一家代理。现在程序做执行，我出判断：建广告、拉数据、盯效果、写报告都自动化，你不付工资，不付代理抽成。This used to take 3–5 people plus an agency. Now software executes and I judge: building ads, pulling data, monitoring, reporting are automated. No payroll, no agency cut.
五个环节，每天跑一遍。第 4 步是人；其余由程序完成。Five steps, run daily. Step 4 is a person; the rest is software.
示意数据。流程、口径与规则为实际生产逻辑。Illustrative numbers. The pipeline, metrics and rules are the real production logic.
平台既是运动员又是裁判，它报的转化既有遗漏又有延迟（实测过遗漏近三分之一）。打分一律用你后台的真实数据，平台数据只用来定位细节。The platform is both player and referee; its reported conversions are incomplete and delayed (I've measured nearly a third missing). Scoring uses your backend; platform numbers only locate details.
广告带来的新用户和自己回来的老用户混在一本账里必然误判。两本账各说各的事。Mixing ad-driven new users with organically returning ones misattributes credit in one direction or the other. Two ledgers.
用户付费需要时间，昨天带来的用户，收入要几天到几周才回完。未成熟的日期一律标注排除。Payments take days or weeks to complete. Unfinished dates are flagged and excluded.
单日收入常被个别大额用户左右。按周判断，或者看更早更稳的信号：付费人数比付费金额更早说真话。One day's revenue is often swung by a single big spender. Judge by the week, or by earlier, steadier signals: how many pay tells the truth sooner than how much.
系统只提建议、等确认，人点头后才执行，每一步留记录。这是硬边界。The system proposes and waits. It acts after a person approves, and every step is logged. A hard boundary.
几十个账户、多个时区、多条业务线，靠命名规范和程序核对。看错账户、比错时区、混错业务线这类事故，恰恰是人手多的团队最常犯的。Dozens of accounts across time zones and business lines, reconciled by naming conventions and code. Wrong-account and wrong-timezone mistakes are what bigger teams make most.
| 传统投放团队（3–5 人）Traditional UA team (3–5) | 这套方式（1 人 + AI）This setup (1 person + AI) | |
|---|---|---|
| 分析密度Analysis density | 周报级，抽样看重点Weekly, sampled | 日度全量，每条广告每天过一遍Daily, every ad |
| 信哪个数据Source of truth | 广告平台报的数字，易被假信号带偏The platform's numbers; false signals mislead | 你后台的真实数据，三方对账Your backend, three-way reconciled |
| 经验怎么落地How experience lands | 资深判断经新人之手执行，逐层稀释Senior judgment diluted through junior hands | 判断写成规则直接执行Judgment codified as rules and executed |
| 数据追踪基建Tracking infrastructure | 通常没人会搭，外包或缺失Usually outsourced or absent | 多次从 0 到 1 搭建，随交付Built from zero repeatedly; included |
| 成本结构Cost structure | 3–5 人工资 + 代理服务费（消耗的 5–15%）3–5 salaries + agency fee (5–15% of spend) | 单人顾问费用，不抽消耗One advisor's fee; no cut of spend |
这套方式每天怎么保证 AI 记得在管什么、改过什么、上次改动有没有用，见 运行机制。How the AI is kept aware of what's live, what changed and whether it worked: see Mechanics.
从 0 起盘到大规模放大都亲手做过，$1B+ 累计消耗练出来的判断：什么时候加、什么时候停、什么信号是假警报。From cold start to full scale, hands-on. Judgment built on $1B+ of real spend: when to push, when to stop, which signals are false alarms.
程序建广告、每天全量看数、自动出报告，团队级产出、单人成本。这套工作流也可以落进你自己的业务：数据、运营、报表都能这样做。Ads built by code, every number reviewed daily, reports generated automatically — a team's output at one person's cost. The same workflow can be installed in your own business: data, operations, reporting.
操盘过豆包这样的大众级 AI 产品，也做过多款头部流水游戏的增长。大预算怎么打、小预算怎么省，都做过。Ran growth for Doubao, a mass-market AI product, and for multiple top-grossing games. Large budgets and lean tests, both.
多次从 0 到 1 搭建归因与数据追踪体系。投放的前提是账算得清，这层地基我自己搭，搭好就是你的资产。Attribution and tracking built from zero, repeatedly. Paid growth only works when the math is trustworthy; I lay that foundation myself, and it stays yours.