7 年海外增长，累计操盘媒体消耗 $1 Billion+。
现在，我用媒体 API + AI，一个人交付一个投放团队的完整链路。
Seven years in global growth, $1B+ cumulative media spend managed.
Today I deliver a full UA team's output — one person, media APIs, and AI.
从腾讯到现职，海外 Google、Meta、TikTok、Bing 与国内主流渠道，App 与 PC/Web 双端都完整操盘过；把归因系统、投放链路、媒体采买做成一个整体是我的方法论。其中 PC/Web 端是市面上最稀缺的能力（多数投放人才只懂移动端买量）——恰好也是 AI 产品的主战场。 From Tencent to today: every major channel — Google, Meta, TikTok, Bing — across both App and PC/Web. My method is making attribution, funnel, and paid media work as one system. PC/Web expertise is the market's scarcest skill (most UA talent is mobile-only) — and it happens to be exactly where AI products live.
用户增长（UA，也叫买量），说白了就是：花钱在 Google、Meta、TikTok 这些平台投广告，把你的目标用户买进来，变成注册和付费。它和做内容、等口碑最大的区别只有一条：今天投钱，明天就有数据；一旦算清楚「花 1 块能回来多少」，它就是一台可以不断加码的机器。 User acquisition — paid growth — in plain words: you pay Google, Meta, or TikTok to put your product in front of the right people, and turn them into signups and paying users. Its one decisive difference from content and word-of-mouth: spend today, see data tomorrow — and once you know what a dollar in brings back, it becomes a machine you can keep feeding.
这是买量最迷人的地方：同一套验证过的打法，能在预算、渠道、市场三个维度同时放大——这几乎是商业世界里唯一「验证过就能加码」的获客方式。This is the magic of paid growth: one proven playbook scales along three axes at once — budget, channel, market. Almost nothing else in business lets you double down the moment it's proven.
不达标就放大，等于放大亏损。会踩油门的人到处都是，知道什么时候踩刹车的人才值钱。Scaling an unproven machine just scales the losses. Anyone can hit the gas — knowing when to brake is what's rare.
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」，并告诉你先修什么。Whether your product should spend right 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。就算未来不再合作，这套基建也是你的资产。Tracking and reconciliation are built on your own accounts. Even if we part ways, the infrastructure 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，变成一门可以加码的生意。Once a capped test proves ROI, we scale along budget, channel, and market — customer acquisition stops being luck and becomes a business you can double down on.
投广告这件事，老板最怕的从来是三个问题：钱花出去到底赚没赚回来？数据这么乱信哪个？养团队请代理这么贵值不值？下面按顺序回答，不用任何广告行业黑话。 If you're paying for ads, three questions keep you up at night: is the money actually coming back? Which numbers can I trust? And is a team or agency really worth the cost? Here are the answers, in plain language — no ad-industry jargon.
我把广告数据和你自己后台的真实注册、付费数据接在一起算账，而不是信广告平台自己汇报的成绩单。你第一次能明确回答：这个渠道到底赚不赚钱。I wire ad data to your own backend — real signups, real payments — instead of trusting the ad platform's self-graded report card. For the first time, you can answer: is this channel actually profitable?
系统每天把每一条广告都过一遍：哪条在亏钱、哪条在赚钱、哪条数据反常。反常的先交叉查证再动手——既不放过真问题，也不被假警报吓得乱调。Every single ad gets reviewed every single day: what's losing, what's earning, what looks off. Anomalies get cross-checked before anything moves — real problems get caught, false alarms don't trigger panic.
这些活过去需要 3–5 人团队加一家代理公司。现在系统干执行，我出判断：建广告、拉数据、盯效果、写报告全部自动化，你不发工资、不付代理抽成。This used to take a 3–5 person team plus an agency. Now the system executes and I judge: building ads, pulling data, monitoring, reporting — all automated. No payroll, no agency cut.
五个环节连成一个每天循环的闭环。橙色那一环是我——所有花钱的决定都停在人这里，其余环节由程序完成。 Five links forming a loop that runs daily. The amber link is me — every spending decision stops at a human; everything else is done by software.
投放计划和广告素材自动读取，不靠来回传文件Plans and creatives are read in automatically — no file ping-pong
广告由程序批量创建和修改，不用人在后台一条条点，快且不出错Ads are created and edited programmatically — no console clicking, fast and error-free
广告平台报的 × 你后台真实发生的 × 用户进网站后的每一步，三边对齐才算数What platforms claim × what your backend records × what users actually do on-site — numbers count only when all three agree
加预算、砍预算、停广告——花钱的决定 100% 由人做Spend more, spend less, stop — every money decision is made by a human
决定由程序执行，每一步有记录可查；报告自动发到你手上Decisions executed by code, every step logged; reports land in your inbox automatically
△ 演示数据。流程、口径与规则均为真实生产逻辑。△ Illustrative data — the pipeline, calibers, and rules are the real production logic.
平台既是运动员又是裁判：它汇报的「转化」既有遗漏又有延迟（我实测过遗漏近三分之一的情况）。打分一律用你自己后台的真实数据，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 going missing). Scoring always runs on your own backend data — platform numbers only help locate details.
广告拉来的新用户，和自己回来的老用户，混在一本账里必然误判：要么把产品和运营的功劳算给广告，要么反过来。我给你两本账，各说各的事。New users brought by ads and old users coming back on their own must not share one ledger — you'd credit ads for the product's work, or the reverse. You get two ledgers, each telling its own truth.
用户付费需要时间：昨天拉来的用户，收入要陆续几天甚至几周才回完。拿「还没回完账」的日期说广告亏了，会砍掉一条正在赚钱的广告——没到期的日期一律标注排除。Payments take time: users acquired yesterday keep paying over days or weeks. Judge an unfinished date and you'll kill an ad that's actually earning — unfinished dates are flagged and excluded.
单日收入常被个别大额客户左右，大起大落是常态不是信号。按周判断，或者看更早、更稳的信号：付费的「人数」比付费的「金额」更早说真话。One day's revenue is often swung by a single big spender — swings are noise, not signal. Judge by the week, or by earlier, steadier signals: how many people pay tells the truth sooner than how much they pay.
加预算、砍预算、停广告——系统只能提建议、等确认，人点头后才执行，且每一步留有记录。这是硬边界，没有例外。Raising budgets, cutting them, stopping ads — the system may only propose and wait. It acts after a human nod, and every step is logged. A hard boundary, no exceptions.
几十个账户、多个时区、多条业务线，全靠命名规范和程序自动核对。杜绝看错账户、比错时区、混错业务线——这类人为事故，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. No wrong-account, wrong-timezone, wrong-line mistakes — precisely the human errors bigger teams make most.
| 传统投放团队（3–5 人）Traditional UA team (3–5 people) | 这套系统（1 人 + AI）This system (1 person + AI) | |
|---|---|---|
| 分析密度Analysis density | 周报级、抽样看重点Weekly reports, sampled highlights | 日度全量，每条 campaign 每天过一遍Daily and exhaustive — every campaign, every day |
| 信哪个数据Which numbers to trust | 信广告平台报的成绩，易被假信号带偏Trusts the platform's report card; false signals mislead | 以你后台真实数据为准，三方对账Your backend is the truth; three-way reconciliation |
| 经验怎么落地How experience lands | 资深的判断经新人的手执行，逐层稀释Senior judgment diluted through junior hands | 判断写成规则直接执行，零损耗Judgment codified as rules — executed with zero loss |
| 数据追踪基建Tracking infrastructure | 通常没人会搭，外包或干脆裸奔Usually nobody can build it — outsourced, or simply absent | 多次从 0 到 1 搭建经验，自带交付Built from scratch 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. |
从 0 起盘到大规模放大都亲手做过，$1B+ 累计消耗是用真金白银练出来的判断——什么时候该加码、什么时候该刹车、什么信号是假警报，不是理论，是肌肉记忆。From cold start to full scale, all hands-on. $1B+ in cumulative spend forged judgment with real money — when to push, when to brake, which signals are false alarms. Not theory. Muscle memory.
程序建广告、每天全量看数、自动出报告——团队级的产出，单人的成本。而且不止用在投放：这套 AI 工作流我可以手把手帮你落进自己的业务，数据、运营、报表都能这么提效。Ads built by code, every number reviewed daily, reports auto-generated — a team's output at one person's cost. And it's not just for ads: I can implement this AI workflow inside your own business — data, operations, reporting, all accelerated the same way.
操盘过豆包这样的国民级 AI 产品，也做过多款头部流水游戏的增长。大盘子怎么打、小预算怎么省，两头都见过、都做过。Ran growth for Doubao — a household-name AI product — and for multiple top-grossing games. Big budgets and lean tests: seen both, done 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——这层地基我亲手搭过多次，不依赖外包，搭好之后就是你的资产。I've built attribution and tracking systems from scratch, repeatedly. Paid growth only works when the math is trustworthy — I lay that foundation myself, no outsourcing, and it stays as your asset.