npx skills add https://github.com/wondelai/skills --skill cro-methodology基于 CRE 方法论™的科学、以客户为中心的转化率优化方法。非凡的改进来自于理解访客为何不转化,而不是复制竞争对手或应用通用技巧。
不要猜测——去发现。 该方法论拒绝"最佳实践"和"神奇按钮",支持基于证据的优化。大多数网站表现不佳不是因为设计糟糕,而是因为没有人系统地研究过访客为何在未转化的情况下离开。
基础: 每个未转化的访客都有一个原因。你的工作是通过研究去发现这些原因,然后系统地用证据和证明来消除它们。这种以客户为中心的方法始终优于直觉、竞争对手复制和"专家"意见。
目标:10/10。 在审查或创建落地页、漏斗或转化流程时,根据以下原则的遵守程度进行 0-10 分评分。10/10 表示完全符合所有准则;较低的分数表示需要解决的差距。始终提供当前分数以及达到 10/10 所需的具体改进措施。
核心概念: 一个系统化的 9 步流程,用于优化转化率,从定义成功指标开始,经过研究、实验,并将成功经验扩展到整个业务。
为何有效: 随机的优化努力之所以失败,是因为它们跳过了关键的研究步骤。CRE 流程迫使你在更改任何内容之前先了解访客,确保更改基于证据而非观点。
关键见解:
产品应用:
| 上下文 | CRO 流程步骤 | 示例 |
|---|
广告位招租
在这里展示您的产品或服务
触达数万 AI 开发者,精准高效
| 落地页审核 | 步骤 1-3:定义目标、绘制漏斗图、研究访客 | 识别出 70% 的流量跳出是因为价值主张不明确 |
| 结账优化 | 步骤 2:绘制漏斗图以查找阻塞的动脉 | 发现运费冲击导致 40% 的购物车放弃 |
| 新功能发布 | 步骤 6-8:制定策略、设计、实验 | 在全面推广之前 A/B 测试两种定位方法 |
| 电子邮件序列 | 步骤 9:扩展成功经验 | 将落地页中成功的异议处理文案应用到滴灌邮件中 |
| 竞争对手响应 | 步骤 4:市场情报 | 从相邻行业转移经过验证的策略 |
文案模式:
道德边界: 切勿操纵测试结果或选择性使用数据。报告所有测试,包括失败的测试,并等待真正的统计显著性。
详见:testing-methodology.md 以获取详细的 ICE 评分、A/B 测试与多变量测试指南以及统计严谨性说明。
核心概念: 访客不转化有具体的、可发现的原因。研究方法——退出调查、聊天记录、支持工单、销售电话、评论——揭示了"客户之声"及其真实的异议。
为何有效: 公司猜测访客离开的原因,但猜测几乎总是错误的。直接研究始终能发现团队从未预料到的异议,而且客户使用的语言比任何文案撰稿人的发明更具说服力。
关键见解:
产品应用:
| 上下文 | 研究方法 | 示例 |
|---|---|---|
| 退出意图 | 站内调查(Hotjar, Qualaroo) | "是什么阻止您今天注册?" |
| 购买后 | 7 天内的电子邮件调查 | "是什么差点阻止您购买?" |
| 异议挖掘 | 支持工单分析 | 搜索"但是"、"然而"、"担心"等模式 |
| 客户之声 | 销售电话录音 | 捕捉客户用来描述问题的确切语言 |
| 竞争差距 | 评论挖掘(您自己和竞争对手的) | 负面评论 = 未解决的异议 |
文案模式:
道德边界: 在研究过程中尊重客户隐私。对数据进行匿名化处理,获取录音许可,不要过于激进地进行调查,以免降低用户体验。
详见:RESEARCH.md 以获取工具、调查问题和数据分析方法。
核心概念: 每家公司都有被忽视的证明元素——未展示的推荐信、未提及的奖项、未突出的统计数据、不显眼的保证、隐藏的团队资质。这些都是必须清点、获取和展示的"说服力资产"。
为何有效: 访客基于证据和证明做出决定,而非声明。没有证明的大胆声明只是噪音。有充分证明的适度声明则令人无法抗拒。大多数公司坐拥从未使用过的证明金矿。
关键见解:
产品应用:
| 上下文 | 说服力资产 | 示例 |
|---|---|---|
| 落地页页眉 | 徽标栏 + 评分 | "受 10,000+ 家公司信赖"并附上 5 个可识别的徽标 |
| 定价页面 | 风险逆转 | "30 天无理由退款保证" |
| 产品页面 | 具体推荐信 | 照片 + 姓名 + 公司 + "在 3 周内将转化率提高了 47%" |
| 结账流程 | 表单附近的信任徽章 | 安全认证、支付徽标、保证印章 |
| 关于页面 | 团队资质 | 多年经验、认证、出版物、专利 |
文案模式:
道德边界: 切勿捏造推荐信、夸大统计数据或展示虚假的信任徽章。所有证明必须是真实且可验证的。
详见:PERSUASION.md 以获取完整的说服力资产清单和心理触发因素。
核心概念: 异议/反异议(O/CO)表格是核心的 CRE 技术。创建一个两列表格,将每个访客异议映射到具体的、有证据支持的反异议。
为何有效: 访客带着异议而来。如果页面没有解决这些异议,访客就会离开。O/CO 框架确保没有异议得不到回应,并且反异议被精确地放置在阅读流程中异议自然出现的位置。
关键见解:
产品应用:
| 上下文 | 异议类型 | O/CO 示例 |
|---|---|---|
| 信任 | "我为什么要相信你?" | 具体推荐信、媒体徽标、奖项、退款保证 |
| 价格 | "它值这个价吗?" | 投资回报率计算器、与替代方案的成本比较、付款计划 |
| 匹配度 | "它适用于我的情况吗?" | 来自类似客户的案例研究、细分落地页、免费试用 |
| 时机 | "为什么我现在就要行动?" | 延迟成本计算、真正的限时优惠、季节性相关性 |
| 努力程度 | "这会有多难?" | "为您完成"的框架、"5 分钟设置完成"、分步说明 |
文案模式:
道德边界: 用诚实的反异议解决真实的异议。切勿忽视合理的担忧或使用欺骗手段来克服有效的犹豫。
详见:OBJECTIONS.md 以获取完整的 O/CO 框架、研究方法和反异议技巧。
核心概念: 每个实验都需要一个记录在案的假设,将特定的更改与预期的结果联系起来,并基于研究给出理由。使用 ICE 评分(影响力、信心、难易度)进行优先级排序。
为何有效: 没有假设,你只是在随机更改事物。假设迫使你阐明更改为什么应该有效,这意味着它必须基于客户研究。ICE 评分可以防止团队在低影响力的"微小调整"上浪费时间。
关键见解:
产品应用:
| 上下文 | 假设示例 | ICE 评分 |
|---|---|---|
| 标题重写 | "如果我们使用调查中的客户语言,转化率将会提高,因为访客看到他们自己的话被反映出来" | I:8, C:9, E:10 = 9.0 |
| 视频推荐信 | "如果我们添加解决价格异议的视频推荐信,注册量将会增加,因为访客需要信任证明" | I:7, C:7, E:6 = 6.7 |
| 结账重新设计 | "如果我们将结账简化为一页,完成率将会提高,因为分析显示第二步有 40% 的流失" | I:9, C:6, E:3 = 6.0 |
| 按钮颜色 | "如果我们将按钮从蓝色改为绿色,点击量将会增加,因为绿色意味着前进" | I:2, C:2, E:10 = 4.7 |
文案模式:
道德边界: 诚实地报告所有测试结果,包括失败的测试。切勿选择性使用数据或一直运行测试直到得到你想要的结果。
详见:testing-methodology.md 以获取 ICE 评分表和详细的优先级排序说明。
核心概念: 运行对照实验,比较页面版本以确定哪个表现更好,使用适当的统计严谨性确保结果是真实的,而非随机噪音。
为何有效: 没有对照实验,你无法区分真正的改进和随机变化。正确的 A/B 测试方法论可以防止最常见的错误:过早查看和停止、样本量不足、忽视实际显著性以及多重比较问题。
关键见解:
产品应用:
| 上下文 | 测试类型 | 示例 |
|---|---|---|
| 概念验证 | A/B 测试(2-4 个变体) | 基于不同的客户洞察,测试两种根本不同的页面布局 |
| 元素优化 | 多变量测试(10 万+ 访客) | 在已证明的获胜页面上测试 3 个标题 x 3 张图片 x 2 个行动号召 |
| 低流量 | 大胆的 A/B 测试 | 进行可通过较小样本检测到的巨大更改(约 4,000 名访客以检测 50% 的提升) |
| 高流量 | 快速迭代 | 在非重叠页面上并行运行测试,每月 10-20 个测试 |
| 测试后 | 扩展成功经验 | 将成功的洞察应用到落地页、广告文案、电子邮件序列中 |
文案模式:
道德边界: 切勿操纵统计方法来制造显著性。诚实地报告置信区间,并在结果不确定时予以承认。
详见:testing-methodology.md 以获取统计显著性、样本量计算和平台比较信息。
| 错误 | 为何失败 | 解决方法 |
|---|---|---|
| 盲目复制竞争对手 | 你不知道他们的方法是否对他们有效,更不用说对你了 | 研究你的访客的异议,并建立你的证据 |
| 在理解异议之前测试按钮颜色 | 解决表面症状,而非根本原因;微小的效果浪费样本量 | 先进行客户研究,然后根据发现测试大的更改 |
| 假设你知道访客离开的原因 | 团队对访客动机的猜测几乎总是错误的 | 使用退出调查、聊天记录和支持分析来发现真实原因 |
| 使用未经验证的"最佳实践" | 在其他地方有效的方法可能不适用于你的受众、产品或背景 | 将最佳实践视为需要测试的假设,而非必须遵循的规则 |
| 基于 HiPPO 做决策 | 最高薪者的意见不是数据;权威偏见扼杀优化 | 让研究和测试结果决定更改,而不是资历 |
| 在没有漏斗上下文的情况下优化页面 | 改进一个步骤可能会将问题转移到另一个步骤;错过最大的机会 | 首先绘制整个漏斗图,识别阻塞的动脉,按影响力排序 |
| 进行"微小调整"而非大胆更改 | 小的更改很少能达到统计显著性;浪费时间和流量 | 测试可能使转化率翻倍的更改,而不是仅推动 2% |
| 一次失败测试后就放弃 | 机会仍然存在;你只是还没有找到解决方案 | 调查原因,回到研究阶段,尝试更大胆的更改 |
审核任何落地页或转化流程:
| 问题 | 如果答案为"否" | 行动 |
|---|---|---|
| 我们是否知道访客在此页面上应该采取的唯一行动? | 页面缺乏重点,访客感到困惑 | 定义单一的主要转化目标并移除竞争性的行动号召 |
| 我们是否研究过访客不转化的原因(而非猜测)? | 优化基于假设,而非证据 | 运行退出调查、分析聊天记录、审查支持工单 |
| 我们是否有将异议映射到反异议的 O/CO 表格? | 访客的异议在页面上得不到回应 | 根据研究构建 O/CO 表格,将反异议放置在摩擦点 |
| 价值主张在 5 秒内是否清晰明了? | 访客在理解报价之前就跳出了 | 运行 5 秒测试,使用客户语言重写标题 |
| 说服力资产是否可见(推荐信、奖项、保证)? | 页面做出声明但没有证明,访客不相信 | 审核说服力资产,获取缺失的资产,并显著展示 |
| 我们是否绘制了完整的漏斗图并识别了阻塞的动脉? | 优化了错误的页面或错过了最大的机会 | 绘制每个阶段的流量图,与基准进行比较,按影响力排序 |
优化任何页面时:
此技能基于转化率专家开发的 CRE 方法论™。要了解完整的方法论、详细案例研究和高级技巧,请阅读原书:
Dr. Karl Blanks 和 Ben Jesson 是转化率专家(CRE)的联合创始人,该公司是全球领先的转化率优化专业机构。他们的客户包括 Google、Apple、Amazon、Facebook、Dropbox 和许多其他技术领导者。CRE 的方法论获得了女王企业奖(创新奖),这是英国最高的商业荣誉。Blanks 拥有用户体验博士学位,此前曾在惠普管理可用性研究团队。Jesson 的背景是直复营销和网络开发。他们共同开发了 CRE 方法论,该方法已应用于数百个网站,并持续带来了显著的转化率提升。他们的著作《Making Websites Win》将这一方法论提炼为一个系统的、可重复的、基于证据的网站优化流程。
每周安装次数
678
代码库
GitHub 星标数
272
首次出现
2026 年 1 月 29 日
安全审计
安装于
opencode608
gemini-cli592
codex589
github-copilot560
claude-code558
cursor544
Scientific, customer-centric approach to conversion rate optimization based on the CRE Methodology(TM). Extraordinary improvements come from understanding WHY visitors don't convert, not from copying competitors or applying generic tips.
Don't guess -- discover. The methodology rejects "best practices" and "magic buttons" in favor of evidence-based optimization. Most websites underperform not because of bad design, but because no one has systematically researched why visitors leave without converting.
The foundation: Every visitor who doesn't convert has a reason. Your job is to discover those reasons through research, then systematically eliminate them with evidence and proof. This customer-centric approach consistently outperforms intuition, competitor copying, and "expert" opinions.
Goal: 10/10. When reviewing or creating landing pages, funnels, or conversion flows, rate them 0-10 based on adherence to the principles below. A 10/10 means full alignment with all guidelines; lower scores indicate gaps to address. Always provide the current score and specific improvements needed to reach 10/10.
Core concept: A systematic 9-step process for optimizing conversion rates, moving from defining success metrics through research, experimentation, and scaling wins across the business.
Why it works: Random optimization efforts fail because they skip the critical research steps. The CRE process forces you to understand visitors before changing anything, ensuring changes are based on evidence rather than opinion.
Key insights:
Product applications:
| Context | CRO Process Step | Example |
|---|---|---|
| Landing page audit | Steps 1-3: Define goals, map funnel, research visitors | Identify that 70% of traffic bounces because value prop is unclear |
| Checkout optimization | Step 2: Map funnel for blocked arteries | Discover shipping cost shock causes 40% cart abandonment |
| New feature launch | Steps 6-8: Strategize, design, experiment | A/B test two positioning approaches before full rollout |
| Email sequence | Step 9: Scale wins | Apply winning objection-handling copy from landing page to drip emails |
| Competitor response | Step 4: Market intelligence | Transfer proven strategies from adjacent industries |
Copy patterns:
Ethical boundary: Never manipulate test results or cherry-pick data. Report all tests, including failures, and wait for genuine statistical significance.
See: testing-methodology.md for detailed ICE scoring, A/B vs. multivariate guidance, and statistical rigor.
Core concept: Visitors don't convert for specific, discoverable reasons. Research methods -- exit surveys, chat logs, support tickets, sales calls, reviews -- reveal the "voice of the customer" and their real objections.
Why it works: Companies guess why visitors leave, but guesses are almost always wrong. Direct research consistently uncovers objections that teams never anticipated, and the language customers use is more persuasive than any copywriter's invention.
Key insights:
Product applications:
| Context | Research Method | Example |
|---|---|---|
| Exit intent | On-site survey (Hotjar, Qualaroo) | "What's preventing you from signing up today?" |
| Post-purchase | Email survey within 7 days | "What almost stopped you from buying?" |
| Objection mining | Support ticket analysis | Search for "but", "however", "worried about" patterns |
| Voice of customer | Sales call recordings | Capture exact language customers use to describe problems |
| Competitive gaps | Review mining (yours and competitors') | Negative reviews = unaddressed objections |
Copy patterns:
Ethical boundary: Respect customer privacy in research. Anonymize data, get consent for recordings, and don't survey so aggressively that you degrade the user experience.
See: RESEARCH.md for tools, survey questions, and data analysis methods.
Core concept: Every company has overlooked proof elements -- testimonials not displayed, awards not mentioned, statistics not highlighted, guarantees not prominent, team credentials hidden. These are "persuasion assets" that must be inventoried, acquired, and displayed.
Why it works: Visitors make decisions based on evidence and proof, not claims. A bold claim without proof is just noise. A modest claim with overwhelming proof is irresistible. Most companies sit on goldmines of proof they never use.
Key insights:
Product applications:
| Context | Persuasion Asset | Example |
|---|---|---|
| Landing page header | Logo bar + rating | "Trusted by 10,000+ companies" with 5 recognizable logos |
| Pricing page | Risk reversal | "30-day money-back guarantee, no questions asked" |
| Product page | Specific testimonial | Photo + name + company + "Increased conversion by 47% in 3 weeks" |
| Checkout flow | Trust badges near forms | Security certification, payment logos, guarantee seal |
| About page | Team credentials | Years of experience, certifications, publications, patents |
Copy patterns:
Ethical boundary: Never fabricate testimonials, inflate statistics, or display fake trust badges. All proof must be genuine and verifiable.
See: PERSUASION.md for the full persuasion assets checklist and psychological triggers.
Core concept: The Objection/Counter-Objection (O/CO) table is the core CRE technique. Create a two-column table mapping every visitor objection to specific, evidence-backed counter-objections.
Why it works: Visitors arrive with objections. If the page doesn't address them, visitors leave. The O/CO framework ensures no objection goes unanswered, and counter-objections are placed exactly where objections naturally arise during the reading flow.
Key insights:
Product applications:
| Context | Objection Type | O/CO Example |
|---|---|---|
| Trust | "Why should I believe you?" | Specific testimonials, media logos, awards, money-back guarantee |
| Price | "Is it worth the money?" | ROI calculator, cost comparison vs. alternatives, payment plans |
| Fit | "Will it work for MY situation?" | Case studies from similar customers, segmented landing pages, free trial |
| Timing | "Why should I act now?" | Cost of delay calculation, genuine limited-time offers, seasonal relevance |
| Effort | "How hard will this be?" | "Done for you" framing, "Set up in 5 minutes", step-by-step breakdown |
Copy patterns:
Ethical boundary: Address real objections with honest counter-objections. Never dismiss legitimate concerns or use deception to overcome valid hesitations.
See: OBJECTIONS.md for the full O/CO framework, research methods, and counter-objection techniques.
Core concept: Every experiment needs a documented hypothesis linking a specific change to an expected outcome with a reason grounded in research. Prioritize using ICE scoring (Impact, Confidence, Ease).
Why it works: Without a hypothesis, you're just changing things randomly. The hypothesis forces you to articulate WHY a change should work, which means it must be grounded in customer research. ICE scoring prevents teams from wasting time on low-impact "meek tweaks."
Key insights:
Product applications:
| Context | Hypothesis Example | ICE Score |
|---|---|---|
| Headline rewrite | "If we use customer language from surveys, conversion will increase because visitors see their own words reflected" | I:8, C:9, E:10 = 9.0 |
| Video testimonial | "If we add video testimonial addressing price objection, signups will increase because visitors need trust proof" | I:7, C:7, E:6 = 6.7 |
| Checkout redesign | "If we simplify checkout to one page, completion will increase because analytics show 40% drop at step 2" | I:9, C:6, E:3 = 6.0 |
| Button color | "If we change button from blue to green, clicks will increase because green means go" | I:2, C:2, E:10 = 4.7 |
Copy patterns:
Ethical boundary: Report all test results honestly, including failures. Never cherry-pick data or run tests until you get the result you want.
See: testing-methodology.md for ICE scoring tables and detailed prioritization.
Core concept: Run controlled experiments comparing page versions to determine which performs better, using proper statistical rigor to ensure results are real, not random noise.
Why it works: Without controlled experiments, you can't distinguish real improvements from random variation. Proper A/B testing methodology prevents the most common errors: peeking and stopping early, insufficient sample size, ignoring practical significance, and the multiple comparison problem.
Key insights:
Product applications:
| Context | Test Type | Example |
|---|---|---|
| Concept validation | A/B test (2-4 variants) | Test two fundamentally different page layouts based on different customer insights |
| Element optimization | Multivariate (100k+ visitors) | Test 3 headlines x 3 images x 2 CTAs on proven winning page |
| Low traffic | Bold A/B test | Make dramatic changes detectable with smaller samples (~4,000 visitors for 50% lift) |
| High traffic | Rapid iteration | Run parallel tests on non-overlapping pages, 10-20 tests/month |
| Post-test | Scale wins | Apply winning insights across landing pages, ad copy, email sequences |
Copy patterns:
Ethical boundary: Never manipulate statistical methods to manufacture significance. Report confidence intervals honestly and acknowledge when results are inconclusive.
See: testing-methodology.md for statistical significance, sample size calculations, and platform comparison.
| Mistake | Why It Fails | Fix |
|---|---|---|
| Copying competitors blindly | You don't know if their approach works for them, let alone for you | Research YOUR visitors' objections and build YOUR evidence |
| Testing button colors before understanding objections | Addresses surface symptoms, not root causes; tiny effects waste sample size | Do customer research first, then test big changes based on findings |
| Assuming you know why visitors leave | Teams are almost always wrong about visitor motivations | Use exit surveys, chat logs, and support analysis to discover real reasons |
| Using "best practices" without validation | What works elsewhere may not work for your audience, product, or context | Treat best practices as hypotheses to test, not rules to follow |
| Making decisions based on HiPPO | Highest Paid Person's Opinion is not data; authority bias kills optimization | Let research and test results determine changes, not seniority |
| Optimizing pages without funnel context | Improving one step may shift problems to another; miss biggest opportunities |
Audit any landing page or conversion flow:
| Question | If No | Action |
|---|---|---|
| Do we know the ONE action visitors should take on this page? | Page lacks focus, visitors are confused | Define single primary conversion goal and remove competing CTAs |
| Have we researched why visitors aren't converting (not guessed)? | Optimization is based on assumptions, not evidence | Run exit surveys, analyze chat logs, review support tickets |
| Do we have an O/CO table mapping objections to counter-objections? | Visitor objections go unanswered on the page | Build O/CO table from research, place counter-objections at friction points |
| Is the value proposition crystal clear within 5 seconds? | Visitors bounce before understanding the offer | Run 5-second test, rewrite headline using customer language |
| Are persuasion assets visible (testimonials, awards, guarantees)? | Page makes claims without proof, visitors don't believe | Audit persuasion assets, acquire missing ones, display prominently |
| Have we mapped the full funnel and identified blocked arteries? | Optimizing wrong page or missing biggest opportunity | Map traffic volume at each stage, compare to benchmarks, prioritize by impact |
When optimizing any page:
This skill is based on the CRE Methodology(TM) developed by Conversion Rate Experts. For the complete methodology, detailed case studies, and advanced techniques, read the original book:
Dr. Karl Blanks and Ben Jesson are the cofounders of Conversion Rate Experts (CRE), the world's leading agency specializing in conversion rate optimization. Their clients have included Google, Apple, Amazon, Facebook, Dropbox, and many other technology leaders. CRE's methodology has been recognized with a Queen's Award for Enterprise (Innovation), the UK's highest business honor. Blanks holds a PhD in user experience and previously managed teams of usability researchers at Hewlett-Packard. Jesson's background is in direct-response marketing and web development. Together they developed the CRE Methodology, which has been applied across hundreds of websites and consistently delivered significant conversion improvements. Their book Making Websites Win distills this methodology into a systematic, repeatable process for evidence-based website optimization.
Weekly Installs
678
Repository
GitHub Stars
272
First Seen
Jan 29, 2026
Security Audits
Gen Agent Trust HubPassSocketPassSnykWarn
Installed on
opencode608
gemini-cli592
codex589
github-copilot560
claude-code558
cursor544
DOCX文件创建、编辑与分析完整指南 - 使用docx-js、Pandoc和Python脚本
41,800 周安装
合作与业务拓展指南:9位产品负责人教你建立战略伙伴关系,达成业务交易
656 周安装
企业级智能体运维指南:云端AI系统生命周期管理、可观测性与安全控制
657 周安装
Amazon Aurora DSQL 技能:无服务器分布式 SQL 数据库管理与 MCP 工具集成
SQL查询优化指南:PostgreSQL、Snowflake、BigQuery高性能SQL编写技巧与方言参考
657 周安装
Gemini Live API 开发指南:实时语音视频交互、WebSockets集成与SDK使用
657 周安装
WordPress渗透测试指南:WPScan工具使用与漏洞扫描实战教程
| Map entire funnel first, identify blocked arteries, prioritize by impact |
| Making "meek tweaks" instead of bold changes | Small changes rarely reach statistical significance; wastes time and traffic | Test changes that could double conversion, not nudge it 2% |
| Giving up after one failed test | The opportunity still exists; you just haven't found the solution yet | Investigate why, go back to research, try a bolder change |