战略与对齐
把愿景转化为可防御的押注、北极星指标和让权衡显式化的 OKR。
Explainer
战略不是愿景声明,不是目标,也不是路线图。它是关于「在哪里玩、如何取胜、以及如何知道自己在赢」这一组连贯选择。多数「战略」演示文档都是穿着战略外衣的志向。真正的战略对那些诱人的方向说不,把团队暴露在能够承受的风险下,并随着时间累积优势。对齐则是把这些选择转化为跨职能连贯行动的过程。
愿景 vs 战略 vs 路线图 vs OKR
这四个词经常被混淆。愿景是你正在创造的长弧未来状态(「每位开发者都能使用可编程的钱」)。战略是在当前约束下如何到达那里的选择(「聚焦原生互联网业务;以 API 质量取胜;按使用量定价;以开发者体验防守」)。路线图是把战略按可交付的赌注排序的表达。OKR 是周期内的承诺,显示赌注是否在奏效。如果你的团队不能把这四者各自塞进一页,你就在混淆它们。
Rumelt 的好战略内核
Richard Rumelt 的三段内核:诊断(真正的问题是什么?)、指导方针(我们的方法是什么?)、连贯行动(具体要做什么?)。多数「战略」是穿着战略外衣的目标清单;它们跳过了诊断(最难的部分),直接进入行动。诊断才是揭示哪些行动连贯、哪些是随机的关键。
- 诊断:一个把复杂性简化、并指出局势中关键面的框架。
- 指导方针:回应诊断、并能创造优势的总体方法。
- 连贯行动:方针所要求的步骤,彼此互相强化而非冲突。
经得起现实检验的产品愿景
强产品愿景描述用户的未来状态,点出用户,并具体到可以让人提出反对。「成为 X 领域的领导者」不是愿景;那是新闻通稿。「任何人都能在不到一小时内从笔记本电脑上把生产应用部署上线,而无需 DevOps 团队」就是愿景:它点出了用户、未来状态、以及隐含的当前痛点。
- 让愿景可证伪:5 年后,它实现了没有?
- 搭配 3-5 年战略和 12 个月路线图;光有愿景是无法发布的。
- 市场转变时重新校准愿景,但抗拒每季度都重写它。
- 在工程师和设计师身上测试 —— 如果不让他们兴奋,它就熬不过 Q3 的疲惫。
Wardley Maps 用于产品战略
Simon Wardley 的地图法把价值链沿一条演化轴绘制(genesis → 定制 → 产品 → 商品)。它显示出你的产品中今天哪些部分是差异化的,哪些即将商品化。战略决策因此变得可见:在哪里投资 genesis、在哪里产品化、在哪里把已商品化的部分外包给 utility。
- X 轴:演化阶段(genesis → 定制 → 产品 → 商品)。
- Y 轴:对用户的可见度(顶部高、底部低)。
- 锚定在用户需求上;依赖向下连成链。
- 组件随时间向右漂移 —— 提前预判,而不是被它惊到。
波特五力(为产品改造)
Porter 的框架 —— 供应商议价能力、买方议价能力、替代品威胁、新进入者威胁、行业内竞争 —— 起初为行业而设,但若把产品表面看成一个行业,它对产品战略同样有效。它能让你在投入某个位置之前,看清竞争结构是友好还是敌对。
- 买方力量:用户切换到替代品(包括什么都不做)有多容易?
- 供应商力量:你的依赖中(云、支付、模型 API)有多少能给你涨价?
- 替代品:有哪些完全不同的方案能解决同一个 job?
- 新进入者:对接下来的 50 家创业公司,你的技术/合规/数据壁垒有多高?
- 行业内竞争:对手是趋同到同一个 playbook,还是在差异化?
北极星指标(NSM)
北极星指标捕捉用户与业务之间的核心价值交换。好的 NSM 是长期留存或营收的领先指标、基于行为、且对产品独有。Amplitude 的框架区分两种 NSM 原型:注意力型产品优化时间/参与,交易型产品优化已完成交易,生产力型产品优化已完成的 job。
- 避免虚荣指标(注册用户、page views)—— 它们会无价值地上升。
- 把 NSM 与团队能直接影响的输入指标配对。
- 在成熟度拐点重新评估 NSM;把你带到 PMF 的指标未必是带你越过 PMF 的那个。
- 没有任何团队能撬动的 NSM 是仪表盘,不是北极星。
OKR vs KPI —— 以及为什么多数 OKR 失败
OKR(Andy Grove → John Doerr)表达的是变化:Objective 是定性愿望,Key Results 是 3-5 个可度量的结果,用以证明 objective 真的发生过。KPI 是健康监测 —— 你看着但不一定要去推动的运营指标。多数 OKR 项目失败,是因为团队对格式照搬而无实质,把输出写成 Key Results,设定灌水或不可能的目标,而且从不在中期复盘。
- Objective 必须鼓舞人心且有时间限制;Key Results 必须可量化、结果二元。
- 上限 3 个 Objective、每个 3-5 个 KR —— 多了就是 wishlist。
- 校准信心:周期开始时预估 50-70% 的信心;若全在 100%,你在灌水。
- 做一次中期检查;不在季度中期审视的 OKR,会沦为季度业绩剧场。
- 周期末按 0-1 打分;0.7 才是目标,而不是 1.0。
产品市场契合(PMF)及如何衡量
PMF 出了名的模糊。两种实操检验:Sean Ellis 的「如果再也不能用这个产品,你会怎样?」(40%+ 选「非常失望」即指示 PMF),以及 Superhuman 引擎,它对「非常失望」的人进行细分,找到你的高期望核心并加倍下注。两者在早期阶段的衡量上都比 NPS 更有效。
- 在 PMF 之前,留存曲线不会变平 —— 每个群组都会向零衰减。
- 在 PMF,留存变平(部分用户会无限期使用),口碑变成有意义的获客渠道。
- 不同产品的 PMF 留存阈值不同;消费者社交 ≠ B2B SaaS ≠ 垂直 SaaS。
- 在没有定性的重复使用证据之前,不要去优化 PMF 指标;你能让问卷动起来,却推不动价值。
市场定位(Positioning)
April Dunford 的框架:定位是定义你的产品「是为谁、属于哪个品类、有什么不同、这种不同对谁重要」的行为。糟糕的定位是隐形的 —— 团队无法一致表达。好的定位是把模糊的信息变成有共鸣的文案、把销售对话变成合格商机的透镜。
- 找出真正的竞争替代方案(常常是电子表格和「什么都不做」)。
- 隔离独有属性 —— 只有你拥有或会做的事。
- 把属性映射到价值:对每一项追问「那又如何?」
- 找出最看重这套组合的买家。
- 选取最能彰显你价值的市场品类。
定价与打包基础
定价是杠杆最大的产品决策之一,也是大多数团队投入最不足的一个。四个杠杆:定价策略(渗透、价值导向、premium)、定价模型(按席位、按用量、分级、混合)、打包(哪些功能进哪个 plan)、折扣政策。Tom Tunguz、Patrick Campbell、Madhavan Ramanujam 一致认为一个核心:按客户的支付意愿定价,而不是按你的成本或竞争对手。
- 在调价前用 Van Westendorp 的四题问卷调查支付意愿。
- 按 segment 分层 —— 企业、中市场、SMB 能承受的价格不同。
- 把功能围绕「good / better / best」打包,锚定在用户 job 的深度上,而不是功能数量上。
- 年付折扣会复利留存;有意识地使用。
- 每年重新定价;价格腐蚀是增长的隐形杀手。
自建 vs 购买 vs 合作
每个 PM 都会面对这个权衡。当能力是核心差异化时,自建;当它是无差异基础设施时(auth、支付、可观测性),购买;当能力位于他人擅长之处、而集成本身就是价值时,合作。经典错误是因为「看起来不难」而从零自建 auth 或支付,然后才发现长尾的合规、边缘情况和维护成本。
- 核心能力 + 差异化执行 → 自建。
- 无差异能力 + 已商品化的供应商 → 购买。
- 邻近能力 + 强势伙伴 → 合作,且对数据与营收边界明确划分。
- 每年重估;市场在演化,昨日的 build 可能成为明日的 buy。
Now / Next / Later 作为路线图格式
Janna Bastow 的 Now / Next / Later 格式,是对基于日期的功能承诺式路线图的解药。「Now」是进行中的;「Next」是为下一周期承诺的;「Later」是方向性、未承诺的。它给团队留出空间去探索与迭代,而不必承担挪动甘特图条的政治成本。
- 把每一项表达为问题主题 + 目标结果,而不是功能名。
- 把 Now 限制在活跃产能;Next 限制在下一个季度;Later 限制为大主题。
- 与 OKR 配套,使主题可以连接到结果。
- 明确传达信心程度:「已承诺」「已规划」「探索中」。
Framework atlas
Reference cards for each method in this mission
Expand a card for when to deploy it, misuse patterns, sequencing guidance, and (where relevant) shorthand formulas.
Strategy formulation · Richard Rumelt, Good Strategy / Bad StrategyRumelt's Kernel
Diagnosis, guiding policy, coherent action. Forces a strategy past 'list of goals' into a defensible logic.
When to use
- Drafting product strategy for a new charter.
- Auditing an existing strategy that 'doesn't feel like one'.
When not to
- Tactical sprint planning where the strategy is already set.
How to apply
- Write the diagnosis as a single paragraph: what is the critical structural feature of this situation?
- Write the guiding policy: in light of that diagnosis, what is our overall approach?
- Write the coherent actions: what specific moves does the policy demand?
- Stress-test by asking: 'what does this strategy explicitly *not* do?'
- Starting with actions and reverse-engineering a diagnosis.
- Diagnoses that are slogans rather than explanations.
Strategic landscape · Simon WardleyWardley Maps
Map of value-chain components on user-visibility (Y) and evolution (X) axes. Reveals where to invest, defend, and commoditize.
When to use
- Long-term platform decisions.
- Build vs buy vs partner debates.
- Understanding why a competitor's approach feels disorienting.
When not to
- Quarterly tactical planning — overhead is too high.
How to apply
- Anchor on a user need at the top of the Y-axis.
- Decompose the value chain downward.
- Place each component on the evolution axis.
- Identify drift: which components are about to commoditize?
- Decide investment: build at genesis, product-ize what's core, outsource the commoditized.
Competitive analysis · Michael Porter (1979)Porter's Five Forces (Adapted)
Industry analysis tool adapted for product surfaces. Identifies the structural forces that determine profitability and competitive intensity.
When to use
- Entering a new product category.
- Sanity-checking why a category leader is profitable while challengers struggle.
When not to
- Tactical positioning of a single feature; this is a whole-market lens.
How to apply
- Score buyer power, supplier power, substitutes, new entrant threat, rivalry on a 1-5 scale.
- Total > 18: hostile market — only enter with a structural advantage.
- Total < 12: friendly market — the opportunity is real, but expect imitators.
Product-Market Fit · Sean Ellis (2009)Sean Ellis PMF Test
Survey users with: 'How would you feel if you could no longer use this product?' Threshold: ≥40% answer 'very disappointed' indicates PMF.
When to use
- Early-stage products trying to verify PMF.
- Re-running after a big pivot.
When not to
- Mature products where the question saturates ('most loyal users always say very disappointed').
How to apply
- Survey active users (not registered users — only those who've used the product recently).
- Three options: very disappointed / somewhat disappointed / not disappointed.
- Segment 'very disappointed' answers by persona, plan, use case.
- Identify your high-expectation customer; double down on what they value.
Roadmapping · Janna Bastow / ProdPadNow / Next / Later
Three-bucket roadmap format with explicit time-horizon and confidence levels.
When to use
- Replacing date-based Gantt-style roadmaps.
- External communication of roadmap to customers without overcommitting.
When not to
- Heavily contractual environments where customers need date commitments.
How to apply
- Define themes around outcomes, not features.
- Bucket as Now (in flight), Next (committed cycle ahead), Later (directional).
- Cap each bucket; if Now overflows, it's not actually 'now'.
- Refresh monthly; reset themes against OKRs each quarter.
Pricing research · Peter van Westendorp (1976)Van Westendorp Price Sensitivity Meter
Four-question pricing battery: at what price would the product be too cheap (suspect quality), cheap (a bargain), expensive, prohibitively expensive? Plotted curves reveal the acceptable range and optimal price point.
When to use
- Pricing for a new product.
- Re-pricing after material changes in scope.
When not to
- Late-stage pricing optimization where A/B tests on real cohorts give better signal.
How to apply
- Survey 200+ qualified prospects.
- Plot the four cumulative curves.
- Identify the Optimal Price Point (intersection of 'too cheap' and 'too expensive').
- Validate with a small cohort experiment before rolling out.
Product Psychology
Cognitive biases that distort product decisions
Availability Cascade
Beliefs gain plausibility through repetition in public discourse, regardless of underlying evidence.
Product Risk
Strategies converge on whatever is fashionable in PM Twitter / podcasts (PLG, AI-everywhere, freemium) without checking fit to the team's specific context.
Research Countermove
Before adopting a popular strategy, write down which structural features of your business *should* make it work; if you can't, you're following a trend.
Strategy as Wishlist Bias
Mistaking a list of objectives for a strategy because both are written in similar language.
Product Risk
Resources spread evenly across many priorities; nothing differentiates; the team is busy and undifferentiated.
Research Countermove
Apply Rumelt's kernel test: can you state the diagnosis and the explicit choices about what *not* to do?
Outcome Halo Bias
Successful companies' visible practices are credited with their success; their unsuccessful peers' identical practices are forgotten.
Product Risk
Adopting Amazon's PR/FAQ, Spotify's tribes, or Netflix's culture deck because they 'work', without controlling for survivorship.
Research Countermove
When borrowing a strategic practice, find at least three failed adopters and understand why; if you can't, you're cargo-culting.
Hierarchical Discounting
Information from senior people is weighted higher than information from people closer to the work.
Product Risk
An exec's hallway comment becomes a quarterly objective; line PMs and engineers' early signals about market reality are ignored.
Research Countermove
Make information sources explicit in strategy docs: cite raw evidence and source level, then weight deliberately rather than by default.
Organizational anti-patterns
When ceremonies look like rigor but aren't
The Strategy-as-Roadmap Antipattern
The 'strategy' deck is a feature timeline. Diagnosis and guiding policy are missing.
Roadmaps feel concrete; strategy feels abstract. Teams skip the hard part.
Insert a one-page strategy doc — diagnosis, policy, choices — between the vision and the roadmap. Validate that every roadmap item ladders to a guiding policy choice.
The Vision Statement That Forgets
The vision is on a slide nobody opens; PMs decide based on stakeholder loudness instead.
The vision is too vague to disqualify any idea, so it disqualifies none.
Sharpen the vision until it can disqualify at least 3 reasonable feature ideas. If it can't, it's not a vision; it's a banner.
OKR Theater
OKRs are written, scored, and forgotten; the same backlog ships regardless of what was committed.
OKRs are imposed top-down without team agency; teams treat them as performance theater.
Co-author OKRs at team level; review at least monthly; allow mid-cycle re-scoping with public rationale.
Sandbagged KRs
Every quarter the team hits 1.0 on every Key Result. Stakeholders smell weakness.
Teams set KRs they're 95% confident of hitting to avoid review-day pain.
Predict pre-quarter confidence (target 60-70%); track confidence over time; reward calibrated honesty over guaranteed wins.
Strategy by Slogan
The strategy fits on a coffee mug ('mobile-first', 'AI-native', 'platform play') but no operational decisions follow.
Slogans are easier to remember and cheaper to defend than coherent action.
For every slogan, write the choice it implies in dollars, headcount, and explicit not-doings.
Worked examples
Walkthroughs translated from real trade-off rooms
Diagnosing the real strategy problem before naming KRs
A B2B startup's revenue is flat. Leadership wants OKRs to 'grow revenue 3x'.
- Resist jumping to KRs. Run a one-page diagnosis: pipeline is healthy, win rate is 22% (industry 35%), churn is 18% annual (industry 8%).
- Identify the real problem: leakage, not acquisition.
- Guiding policy: prioritize retention as the lever for growth this year; pause new logo investment.
- Coherent actions: instrument churn reasons, restructure CS coverage, ship product fixes for the top 3 churn drivers.
- Translate to OKRs: net revenue retention from 84% to 105%, gross churn from 18% to 11%.
TakeawayThe 'grow revenue 3x' goal was symptom-level. The kernel approach revealed retention as the actual leverage.
Picking a North Star Metric for a marketplace
A two-sided marketplace has revenue, GMV, transactions, time-spent, and registered users on its dashboard. The team can't agree on what to optimize.
- Apply Amplitude's archetype: this is a transaction product, not attention.
- Candidate NSMs: GMV, completed transactions, repeat-buyer transactions.
- Test against the criteria: leading? behavioral? unique? — repeat-buyer transactions wins on all three (catches both supply and demand health, leads revenue, reflects value exchange).
- Define input metrics: time-to-first-transaction, supply density per region, repeat purchase rate within 30 days.
TakeawayThe NSM choice was not arbitrary; it had to satisfy three structural criteria. Repeat-buyer transactions captured value-exchange more durably than top-line GMV.
Killing a strategic option using Wardley mapping
A devtools company is debating whether to build its own observability stack.
- Map the value chain anchored on the developer need: detect production issues fast.
- Plot components: alerting (genesis), dashboards (product), log storage (commodity), metric storage (commodity), tracing (between product and commodity).
- Realize storage is firmly commoditized; building it would invest in a commoditizing capability and starve genesis investment.
- Decide: buy storage from a vendor, build the differentiating layer (developer-context-aware alerting and triage).
TakeawayWardley mapping made the commoditization visible early enough to avoid an 18-month, undifferentiated build.
Resources / Case Studies
Curated reading for this mission
Richard Rumelt
The reference text for the kernel of good strategy and the diagnostics of bad strategy.
Distinguishes real strategy from goal-setting; the most useful strategy book a PM can read.
Simon Wardley
Wardley's serial of Medium chapters covering value-chain mapping, the climatic patterns, and doctrines.
The single most powerful strategy tool a PM can learn for free; foundational for build-vs-buy and platform thinking.
April Dunford
Dunford's framework for positioning, complete with the five components and worked examples from B2B SaaS.
The clearest practical book on positioning; PMs and PMMs use the same vocabulary after reading it.
John Cutler & Amplitude
Free 100-page playbook on choosing and operationalizing a North Star Metric.
The most opinionated, practical guide to NSM selection in print; opinionated enough to argue with.
John Doerr
The popularizer book on OKRs with case studies from Google, Intel, and the Gates Foundation.
Lays out the cultural mechanics of OKRs that pure-format guides miss.
A detailed account of using Sean Ellis's PMF survey, segmentation, and feature prioritization to grind toward fit.
Shows how qualitative feedback and quantitative thresholds work together as an operational PMF engine.
Madhavan Ramanujam, Georg Tacke
The Simon-Kucher consultants' framework for designing pricing and packaging into the product itself, not as an afterthought.
The most concrete reference for a PM driving pricing decisions, including survey-based willingness-to-pay research.
Operational playbooks on growth strategy, pricing, retention, and platforms. Paid but the public summaries are useful.
Functional, modern, post-blog-post strategy material from operators currently in the trenches.
Lenny Rachitsky
Recurring case studies on strategy, PMF, and positioning from PMs at major tech companies.
Active, current source of practitioner strategy stories.
Ben Thompson
Daily strategic analysis of the technology industry. Aggregation theory, platform dynamics, and the strategy behind major moves.
Sharpens strategic vocabulary and external framing; PMs who read Stratechery weekly think about competitors and platforms more clearly.
Colin Bryar & Bill Carr
Amazon's mechanisms — PR/FAQ documents, the bar raiser, S-team goals — explained by two former Amazon executives.
Operational details on Amazon's strategic mechanisms, with templates PMs can adapt directly.
Geoffrey Moore
Classic technology adoption lifecycle: innovators → early adopters → chasm → early/late majority. Strategy implications for go-to-market.
The mental model for B2B GTM strategy; even 30 years on, the chasm is still where products die.