[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7E2WGE7T6F1WDKD3M9RT9C8":3,"skill-same:01M7E2WGE7T6F1WDKD3M9RT9C8":49},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":29,"commitSha":30,"skillDirectory":31,"forksCount":14,"repositoryHtmlUrl":32,"repositoryHomepageUrl":32,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":34,"install":35,"resources":48},"01M7E2WGE7T6F1WDKD3M9RT9C8","cro",null,"Use this skill for conversion rate optimization on marketing pages and forms. Trigger phrases include: \"improve our conversion rate,\" \"CRO audit,\" \"why isn't our landing page converting,\" \"analyze our homepage conversions,\" \"optimize our pricing page,\" \"run a CRO analysis,\" \"write a hypothesis for an A\u002FB test,\" \"what should we A\u002FB test,\" \"design an A\u002FB test,\" \"calculate sample size for a test,\" \"find friction on our page,\" \"quick CRO wins,\" or \"optimize our lead capture form.\" This skill applies to homepages, landing pages, pricing pages, and lead capture forms. Distinct from signup flow optimization and in-app paywalls.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fadvertising\u002Fcro\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: cro\ndescription: >\n  Use this skill for conversion rate optimization on marketing pages and forms. Trigger phrases include: \"improve our conversion rate,\" \"CRO audit,\" \"why isn't our landing page converting,\" \"analyze our homepage conversions,\" \"optimize our pricing page,\" \"run a CRO analysis,\" \"write a hypothesis for an A\u002FB test,\" \"what should we A\u002FB test,\" \"design an A\u002FB test,\" \"calculate sample size for a test,\" \"find friction on our page,\" \"quick CRO wins,\" or \"optimize our lead capture form.\" This skill applies to homepages, landing pages, pricing pages, and lead capture forms. Distinct from signup flow optimization and in-app paywalls.\nsource: \"https:\u002F\u002Fgithub.com\u002Fscayver\u002Fmarketing-skills\"\nsource_repository: \"scayver\u002Fmarketing-skills\"\nsource_path: \"skills\u002Fcro\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n# CRO Skill\n\n## Mandatory Content Standards\n\n- Match the output length to the task. For full audits, strategies, plans, or multi-part deliverables, write 1,500 to 10,000 words. For quick tasks, single assets, snippets, or narrow revisions, keep the output concise and provide only the useful variations, rationale, and next steps.\n- Write in a way that sounds like a knowledgeable human wrote it. No robotic or templated phrasing.\n- Use short sentences. One idea per sentence. One focus per paragraph.\n- Use active voice. Never passive constructions.\n- Address the reader directly using \"you\" and \"your.\"\n- Use bullet points only when they genuinely improve readability.\n- Replace all em dashes with commas, parentheses, semicolons, or a new sentence. No hidden Unicode characters.\n- End every sentence with a period.\n- No hashtags, emojis, or asterisks.\n- No introductory or closing filler phrases such as \"in conclusion,\" \"in summary,\" or \"in a world where.\"\n- No warnings, notes, or disclaimers. Stick to requested output.\n- No AI cliches: no \"game-changer,\" \"unlock,\" \"leverage,\" \"dive into,\" \"delve,\" \"cutting-edge,\" \"transformative,\" \"revolutionize.\"\n- No excessive adjectives or adverbs. Let specifics do the work.\n- No broad generalizations. Every claim tied to specific context.\n- Use specific examples, data, and scenarios.\n- Pose at least one thought-provoking question per skill.\n- Mobile-friendly: short paragraphs, clear headers, scannable.\n- Practical and actionable. Every section connects to a next step.\n\n---\n\n## What This Skill Does\n\nThis skill runs conversion rate optimization analysis on marketing pages and forms. The outputs include friction audits, ICE-scored opportunity lists, A\u002FB test hypotheses, test design specifications, sample size calculations, and quick win recommendations.\n\nThis skill applies to:\n\n- Homepages\n- Landing pages (paid and organic)\n- Pricing pages\n- Lead capture and demo request forms\n\nIt does not cover signup flows, onboarding sequences, or in-app upgrade prompts. Those belong in a different scope.\n\nBefore starting a CRO analysis, gather:\n\n- The page URL and current conversion rate (if known)\n- The definition of conversion for this page (trial signup, form submission, demo request, etc.)\n- Monthly unique visitors to the page\n- The primary traffic source (paid, organic, direct, referral)\n- Any existing user research, session recordings, or heatmap data\n- Tools available (Google Analytics, Hotjar, Heap, VWO, Optimizely, or similar)\n- The team's capacity to run and monitor tests\n\n---\n\n## The CRO Mindset\n\nMost teams approach CRO backwards. They start with a change they want to make and then figure out how to justify running a test. The right order is: identify a problem, understand why it exists, form a hypothesis about what would fix it, and then design the test.\n\nThe question that reorients the process: what is the most important reason people who land on this page and have the right intent do not convert?\n\nThat question is never fully answerable from traffic data alone. You need qualitative data (session recordings, user feedback, heatmaps) to understand behavior, and quantitative data (analytics, funnel reports) to know which behaviors are common enough to matter.\n\n---\n\n## Step 1: Define the Conversion Goal and Baseline\n\nBefore you can improve a conversion rate, you need to know what it is and what it means.\n\nIdentify the macro conversion: the primary action the page is designed to drive. For a landing page, it might be form submission. For a pricing page, it might be clicking the primary CTA. For a homepage, it might be navigating to a trial or demo page.\n\nIdentify micro conversions: smaller actions that indicate engagement. Scrolling past the fold, clicking the pricing page link from the homepage, watching a product video, clicking the FAQ. These tell you whether people are engaging with the page even if they do not convert at the macro level.\n\nCalculate the current conversion rate: divide conversions by unique visitors for the same period. If the page gets 5,000 visitors per month and 150 people submit the lead form, the conversion rate is 3%.\n\nSegment the baseline: do not average across traffic sources. A page that converts at 8% for branded search visitors and 1.5% for cold paid social visitors has a very different optimization story than the blended 4% suggests.\n\n---\n\n## Step 2: Full Funnel Analysis\n\nLook at the page as part of a larger funnel, not in isolation. A conversion rate problem on a landing page sometimes originates before the page.\n\nThe questions to answer:\n\nWhere does traffic come from? The intent and expectation of paid search traffic is different from display retargeting traffic. The conversion rate should reflect that difference.\n\nIs there a message match problem? If your ad says \"Free trial, no credit card\" and the landing page leads with a form that requires a credit card, you have a message match failure. Readers feel deceived. They leave. The conversion problem is actually a pre-landing problem.\n\nWhat is the drop-off pattern? Where in the page do people leave? Analytics will tell you scroll depth and exit rate. Session recordings will show you the actual behavior.\n\nWhat is the traffic quality? If conversion rates drop suddenly, check for a change in traffic sources before blaming the page. A landing page that is performing well can look like it is broken if paid traffic shifts from high-intent keywords to broad match.\n\nIs there a technical issue? Slow load time, broken CTAs on mobile, form submission errors, and JavaScript conflicts cause conversions to drop without any copy or design change. Check your page speed score (Google PageSpeed Insights) and your mobile rendering before attributing a drop to anything else.\n\n---\n\n## Step 3: Friction Identification\n\nFriction is anything that makes it harder to convert than it should be. Friction takes many forms.\n\nFriction in the copy:\n\nThe headline does not connect to why the visitor came to the page. The value proposition is buried or absent. Claims without proof. Too much text too early. CTA copy that is generic or unclear.\n\nFriction in the design and UX:\n\nThe primary CTA button is hard to find because it is the same color as the background or positioned below the fold. There is no clear visual hierarchy. The page has competing calls to action. The mobile experience differs significantly from the desktop experience (and mobile traffic is underperforming).\n\nFriction in the form:\n\nToo many fields. Asking for information the visitor does not want to share yet (phone number, company revenue). Confusing field labels. No confirmation of what happens after submission. No clear privacy commitment near the submit button.\n\nFriction in the offer:\n\nThe offer is not clear. The risk feels too high (asking for a 12-month commitment when the visitor does not yet know if the product works). The immediate value of converting is unclear (\"talk to sales\" has less immediate value than \"get your free report now\").\n\nFriction in trust signals:\n\nNo visible social proof. No security badges near sensitive form fields. No recognizable customer names. No clear refund or cancellation policy when money is involved.\n\n---\n\n## Step 4: User Research and Qualitative Data\n\nQuantitative data tells you what is happening. Qualitative data tells you why.\n\nSession recordings (Hotjar, Clarity, FullStory, or similar): Watch 20 to 30 sessions of visitors who did not convert. Look for patterns. Where do they hover and then leave? Which CTAs do they approach and then pull back from? What do they click expecting to go somewhere and find that it does not work?\n\nHeatmaps: Scroll maps show how far people read. Click maps show what they try to interact with. Attention maps show where their eyes spend the most time. Together these reveal which sections are doing work and which are being ignored.\n\nExit surveys: A short one-question survey shown to visitors about to leave the page. \"What stopped you from signing up today?\" Responses are often blunt and specific. Even 50 to 100 responses will reveal common themes.\n\nCustomer interviews: Ask three to five recent signups what they were worried about before they converted and what made them confident enough to proceed. Their language is the language you should use on the page.\n\nOn-site chat transcripts: If you run a chat widget, the questions visitors ask reveal what they could not find or did not understand on the page.\n\n---\n\n## Step 5: ICE Scoring to Prioritize Opportunities\n\nAfter identifying a list of friction points and potential improvements, prioritize them. Do not run every test simultaneously. Tests that overlap contaminate each other.\n\nICE scoring ranks each opportunity on three dimensions:\n\nImpact: If this change worked, how much would it improve the conversion rate? Score 1 to 10.\n\nConfidence: How confident are you that the change will work, based on the evidence? A change backed by session recordings, an exit survey, and a customer interview gets a high confidence score. A change based on a hunch gets a low one. Score 1 to 10.\n\nEase: How much effort does it take to implement and test this change? A headline swap takes hours. A full page redesign takes weeks. Score 1 to 10 (10 being easiest).\n\nCalculate an ICE score by averaging the three dimensions: (Impact + Confidence + Ease) \u002F 3.\n\nSort by ICE score descending and work down the list. Run the highest-scored test first.\n\n---\n\n## Step 6: Hypothesis Writing\n\nA hypothesis is not a guess. It is a structured statement that identifies a problem, proposes a solution, and predicts an outcome.\n\nThe hypothesis format:\n\n\"Because [evidence of a problem], we believe that [proposed change] will [expected outcome] for [audience segment].\"\n\nExample: \"Because 68% of visitors who reach the pricing page exit without clicking any CTA, and session recordings show most users clicking on tier names expecting them to expand with more detail, we believe that adding an expandable feature list to each pricing tier will increase clicks to the trial CTA by at least 15% for visitors who come from organic search.\"\n\nA well-written hypothesis:\n\n- Cites specific evidence (not \"we think people are confused\")\n- Names a specific change (not \"improve the pricing page\")\n- Predicts a specific measurable outcome\n- Identifies the audience segment the test applies to\n\nWrite the hypothesis before designing the test. If you cannot write a clear hypothesis, you do not yet understand the problem well enough to test a solution.\n\n---\n\n## Step 7: A\u002FB Test Design\n\nA well-designed test isolates one variable. You cannot learn from a test that changes the headline, the button color, the copy, and the page layout simultaneously. If it works, you do not know which change drove it. If it fails, you do not know which change killed it.\n\nTest design checklist:\n\nOne variable changed per test. The control is the current page. The variant is the current page with exactly one thing different.\n\nDefine the primary metric before the test runs. Conversion rate for the page's macro conversion goal. Do not add metrics after the test starts; that is called p-hacking and it produces unreliable conclusions.\n\nDefine the secondary metrics. Secondary metrics give context. If conversion rate increases but time on page drops, something unexpected is happening.\n\nSet the test duration before starting. Do not stop a test when it looks like the variant is winning. Let it run to the pre-determined endpoint.\n\nDocument the hypothesis, the variant, and the expected outcome before running. This forces clarity and prevents post-hoc rationalization of results.\n\n---\n\n## Step 8: Sample Size Calculation\n\nRunning a test without enough traffic wastes time and produces unreliable results. Calling a winner with insufficient sample size is a common and costly mistake.\n\nVariables you need to calculate sample size:\n\nBaseline conversion rate: your current conversion rate for this page and goal.\n\nMinimum detectable effect (MDE): the smallest improvement you care about detecting. If your current conversion rate is 3% and you want to detect an improvement to 3.6% (a 20% relative lift), your MDE is 0.6 percentage points.\n\nStatistical significance threshold: typically 95% confidence (meaning you accept a 5% chance of a false positive).\n\nStatistical power: typically 80% (meaning you accept a 20% chance of missing a real effect).\n\nUse a sample size calculator (Optimizely has a free one, as does Evan Miller's online calculator) and input these variables. The output is the number of unique visitors you need in each variant before calling a result.\n\nFor a page with 3% baseline conversion and a 20% MDE, you typically need around 8,000 to 12,000 visitors per variant. At 1,000 visitors per week, that test needs to run for 16 to 24 weeks to reach statistical significance. If your traffic is too low to run a clean test in a reasonable timeframe, do not run the test. Implement the change as a best-bet improvement and measure the before-and-after instead.\n\n---\n\n## Quick Wins: Changes That Rarely Need Testing\n\nSome changes have enough evidence behind them that you can implement them without running a formal test, especially if your traffic volume is too low to support reliable testing.\n\nAdd specificity to the headline. \"Project management for teams\" becomes \"Project management for engineering teams of 20 to 200.\" Specificity increases relevance for the right audience.\n\nRemove phone number field from the lead form. Phone number fields consistently reduce form completion rates by 5 to 20% when they are not necessary for the follow-up process. If your team does not call prospects within 24 hours, remove the field.\n\nMove the primary CTA above the fold. If your main CTA button requires scrolling on most devices, move it up.\n\nAdd a \"no credit card required\" statement near the trial CTA. This single phrase regularly increases trial sign-ups by 5 to 15%.\n\nReplace a generic testimonial with a specific, quantified one. \"Great product!\" does no conversion work. \"We reduced our reporting time from 8 hours per week to 45 minutes\" does.\n\nAdd a comparison table to the pricing page. Buyers who are deciding between your tiers need to know what they gain by upgrading. A visible table with specific feature differences makes that decision easier.\n\nReduce form fields to three or fewer for top-of-funnel offers. Every additional field reduces completion rate. Ask only for what you need to deliver on the offer.\n\n---\n\n## Common CRO Mistakes\n\nTesting without enough traffic: A test that reaches significance at 70% confidence is not reliable. It will produce false positives more often than you want. Wait for 95% confidence.\n\nStopping tests early: Peeking at results and stopping a test when it looks like the variant is winning is called \"optional stopping.\" It dramatically inflates false positive rates. Set your endpoint in advance and stick to it.\n\nRunning too many tests simultaneously: Tests on the same page at the same time contaminate each other unless they are properly isolated (for example, using mutually exclusive audience segments). Run one test on each major conversion point at a time.\n\nIgnoring seasonality: A test that runs from January to February produces different results than the same test run in Q3, because visitor intent and composition change by season. Account for this when interpreting results.\n\nCalling a test based on relative lift alone without checking absolute numbers: A 50% relative improvement from 0.4% to 0.6% conversion is probably within the noise. Look at absolute conversion rate and total conversions, not just relative lift.\n\nNot learning from failed tests: A test that shows no improvement or a decline is not a failure. It is information. Document what you tested, what you expected, what happened, and what you concluded. The best CRO programs are built on a library of this evidence.\n\n---\n\n## Output Formats\n\nFor a CRO audit, deliver:\n\nA friction inventory: a list of every friction point identified, organized by category (copy, design, form, trust, offer), with the evidence behind each one.\n\nAn ICE-scored prioritized opportunity list: the top ten opportunities with scores and rationale.\n\nThree to five fully written hypotheses: in the format specified above, covering the highest-priority opportunities.\n\nSample size calculations for each test: to help the user assess which tests are feasible given their traffic volume.\n\nQuick wins list: changes to implement immediately without waiting for tests.\n\nFor a single A\u002FB test design, deliver:\n\nThe hypothesis, the control description, the variant description, the primary metric, the secondary metrics, the estimated sample size needed, and the recommended test duration based on current traffic.\n","# CRO Skill\n\n## Mandatory Content Standards\n\n- Match the output length to the task. For full audits, strategies, plans, or multi-part deliverables, write 1,500 to 10,000 words. For quick tasks, single assets, snippets, or narrow revisions, keep the output concise and provide only the useful variations, rationale, and next steps.\n- Write in a way that sounds like a knowledgeable human wrote it. No robotic or templated phrasing.\n- Use short sentences. One idea per sentence. One focus per paragraph.\n- Use active voice. Never passive constructions.\n- Address the reader directly using \"you\" and \"your.\"\n- Use bullet points only when they genuinely improve readability.\n- Replace all em dashes with commas, parentheses, semicolons, or a new sentence. No hidden Unicode characters.\n- End every sentence with a period.\n- No hashtags, emojis, or asterisks.\n- No introductory or closing filler phrases such as \"in conclusion,\" \"in summary,\" or \"in a world where.\"\n- No warnings, notes, or disclaimers. Stick to requested output.\n- No AI cliches: no \"game-changer,\" \"unlock,\" \"leverage,\" \"dive into,\" \"delve,\" \"cutting-edge,\" \"transformative,\" \"revolutionize.\"\n- No excessive adjectives or adverbs. Let specifics do the work.\n- No broad generalizations. Every claim tied to specific context.\n- Use specific examples, data, and scenarios.\n- Pose at least one thought-provoking question per skill.\n- Mobile-friendly: short paragraphs, clear headers, scannable.\n- Practical and actionable. Every section connects to a next step.\n\n---\n\n## What This Skill Does\n\nThis skill runs conversion rate optimization analysis on marketing pages and forms. The outputs include friction audits, ICE-scored opportunity lists, A\u002FB test hypotheses, test design specifications, sample size calculations, and quick win recommendations.\n\nThis skill applies to:\n\n- Homepages\n- Landing pages (paid and organic)\n- Pricing pages\n- Lead capture and demo request forms\n\nIt does not cover signup flows, onboarding sequences, or in-app upgrade prompts. Those belong in a different scope.\n\nBefore starting a CRO analysis, gather:\n\n- The page URL and current conversion rate (if known)\n- The definition of conversion for this page (trial signup, form submission, demo request, etc.)\n- Monthly unique visitors to the page\n- The primary traffic source (paid, organic, direct, referral)\n- Any existing user research, session recordings, or heatmap data\n- Tools available (Google Analytics, Hotjar, Heap, VWO, Optimizely, or similar)\n- The team's capacity to run and monitor tests\n\n---\n\n## The CRO Mindset\n\nMost teams approach CRO backwards. They start with a change they want to make and then figure out how to justify running a test. The right order is: identify a problem, understand why it exists, form a hypothesis about what would fix it, and then design the test.\n\nThe question that reorients the process: what is the most important reason people who land on this page and have the right intent do not convert?\n\nThat question is never fully answerable from traffic data alone. You need qualitative data (session recordings, user feedback, heatmaps) to understand behavior, and quantitative data (analytics, funnel reports) to know which behaviors are common enough to matter.\n\n---\n\n## Step 1: Define the Conversion Goal and Baseline\n\nBefore you can improve a conversion rate, you need to know what it is and what it means.\n\nIdentify the macro conversion: the primary action the page is designed to drive. For a landing page, it might be form submission. For a pricing page, it might be clicking the primary CTA. For a homepage, it might be navigating to a trial or demo page.\n\nIdentify micro conversions: smaller actions that indicate engagement. Scrolling past the fold, clicking the pricing page link from the homepage, watching a product video, clicking the FAQ. These tell you whether people are engaging with the page even if they do not convert at the macro level.\n\nCalculate the current conversion rate: divide conversions by unique visitors for the same period. If the page gets 5,000 visitors per month and 150 people submit the lead form, the conversion rate is 3%.\n\nSegment the baseline: do not average across traffic sources. A page that converts at 8% for branded search visitors and 1.5% for cold paid social visitors has a very different optimization story than the blended 4% suggests.\n\n---\n\n## Step 2: Full Funnel Analysis\n\nLook at the page as part of a larger funnel, not in isolation. A conversion rate problem on a landing page sometimes originates before the page.\n\nThe questions to answer:\n\nWhere does traffic come from? The intent and expectation of paid search traffic is different from display retargeting traffic. The conversion rate should reflect that difference.\n\nIs there a message match problem? If your ad says \"Free trial, no credit card\" and the landing page leads with a form that requires a credit card, you have a message match failure. Readers feel deceived. They leave. The conversion problem is actually a pre-landing problem.\n\nWhat is the drop-off pattern? Where in the page do people leave? Analytics will tell you scroll depth and exit rate. Session recordings will show you the actual behavior.\n\nWhat is the traffic quality? If conversion rates drop suddenly, check for a change in traffic sources before blaming the page. A landing page that is performing well can look like it is broken if paid traffic shifts from high-intent keywords to broad match.\n\nIs there a technical issue? Slow load time, broken CTAs on mobile, form submission errors, and JavaScript conflicts cause conversions to drop without any copy or design change. Check your page speed score (Google PageSpeed Insights) and your mobile rendering before attributing a drop to anything else.\n\n---\n\n## Step 3: Friction Identification\n\nFriction is anything that makes it harder to convert than it should be. Friction takes many forms.\n\nFriction in the copy:\n\nThe headline does not connect to why the visitor came to the page. The value proposition is buried or absent. Claims without proof. Too much text too early. CTA copy that is generic or unclear.\n\nFriction in the design and UX:\n\nThe primary CTA button is hard to find because it is the same color as the background or positioned below the fold. There is no clear visual hierarchy. The page has competing calls to action. The mobile experience differs significantly from the desktop experience (and mobile traffic is underperforming).\n\nFriction in the form:\n\nToo many fields. Asking for information the visitor does not want to share yet (phone number, company revenue). Confusing field labels. No confirmation of what happens after submission. No clear privacy commitment near the submit button.\n\nFriction in the offer:\n\nThe offer is not clear. The risk feels too high (asking for a 12-month commitment when the visitor does not yet know if the product works). The immediate value of converting is unclear (\"talk to sales\" has less immediate value than \"get your free report now\").\n\nFriction in trust signals:\n\nNo visible social proof. No security badges near sensitive form fields. No recognizable customer names. No clear refund or cancellation policy when money is involved.\n\n---\n\n## Step 4: User Research and Qualitative Data\n\nQuantitative data tells you what is happening. Qualitative data tells you why.\n\nSession recordings (Hotjar, Clarity, FullStory, or similar): Watch 20 to 30 sessions of visitors who did not convert. Look for patterns. Where do they hover and then leave? Which CTAs do they approach and then pull back from? What do they click expecting to go somewhere and find that it does not work?\n\nHeatmaps: Scroll maps show how far people read. Click maps show what they try to interact with. Attention maps show where their eyes spend the most time. Together these reveal which sections are doing work and which are being ignored.\n\nExit surveys: A short one-question survey shown to visitors about to leave the page. \"What stopped you from signing up today?\" Responses are often blunt and specific. Even 50 to 100 responses will reveal common themes.\n\nCustomer interviews: Ask three to five recent signups what they were worried about before they converted and what made them confident enough to proceed. Their language is the language you should use on the page.\n\nOn-site chat transcripts: If you run a chat widget, the questions visitors ask reveal what they could not find or did not understand on the page.\n\n---\n\n## Step 5: ICE Scoring to Prioritize Opportunities\n\nAfter identifying a list of friction points and potential improvements, prioritize them. Do not run every test simultaneously. Tests that overlap contaminate each other.\n\nICE scoring ranks each opportunity on three dimensions:\n\nImpact: If this change worked, how much would it improve the conversion rate? Score 1 to 10.\n\nConfidence: How confident are you that the change will work, based on the evidence? A change backed by session recordings, an exit survey, and a customer interview gets a high confidence score. A change based on a hunch gets a low one. Score 1 to 10.\n\nEase: How much effort does it take to implement and test this change? A headline swap takes hours. A full page redesign takes weeks. Score 1 to 10 (10 being easiest).\n\nCalculate an ICE score by averaging the three dimensions: (Impact + Confidence + Ease) \u002F 3.\n\nSort by ICE score descending and work down the list. Run the highest-scored test first.\n\n---\n\n## Step 6: Hypothesis Writing\n\nA hypothesis is not a guess. It is a structured statement that identifies a problem, proposes a solution, and predicts an outcome.\n\nThe hypothesis format:\n\n\"Because [evidence of a problem], we believe that [proposed change] will [expected outcome] for [audience segment].\"\n\nExample: \"Because 68% of visitors who reach the pricing page exit without clicking any CTA, and session recordings show most users clicking on tier names expecting them to expand with more detail, we believe that adding an expandable feature list to each pricing tier will increase clicks to the trial CTA by at least 15% for visitors who come from organic search.\"\n\nA well-written hypothesis:\n\n- Cites specific evidence (not \"we think people are confused\")\n- Names a specific change (not \"improve the pricing page\")\n- Predicts a specific measurable outcome\n- Identifies the audience segment the test applies to\n\nWrite the hypothesis before designing the test. If you cannot write a clear hypothesis, you do not yet understand the problem well enough to test a solution.\n\n---\n\n## Step 7: A\u002FB Test Design\n\nA well-designed test isolates one variable. You cannot learn from a test that changes the headline, the button color, the copy, and the page layout simultaneously. If it works, you do not know which change drove it. If it fails, you do not know which change killed it.\n\nTest design checklist:\n\nOne variable changed per test. The control is the current page. The variant is the current page with exactly one thing different.\n\nDefine the primary metric before the test runs. Conversion rate for the page's macro conversion goal. Do not add metrics after the test starts; that is called p-hacking and it produces unreliable conclusions.\n\nDefine the secondary metrics. Secondary metrics give context. If conversion rate increases but time on page drops, something unexpected is happening.\n\nSet the test duration before starting. Do not stop a test when it looks like the variant is winning. Let it run to the pre-determined endpoint.\n\nDocument the hypothesis, the variant, and the expected outcome before running. This forces clarity and prevents post-hoc rationalization of results.\n\n---\n\n## Step 8: Sample Size Calculation\n\nRunning a test without enough traffic wastes time and produces unreliable results. Calling a winner with insufficient sample size is a common and costly mistake.\n\nVariables you need to calculate sample size:\n\nBaseline conversion rate: your current conversion rate for this page and goal.\n\nMinimum detectable effect (MDE): the smallest improvement you care about detecting. If your current conversion rate is 3% and you want to detect an improvement to 3.6% (a 20% relative lift), your MDE is 0.6 percentage points.\n\nStatistical significance threshold: typically 95% confidence (meaning you accept a 5% chance of a false positive).\n\nStatistical power: typically 80% (meaning you accept a 20% chance of missing a real effect).\n\nUse a sample size calculator (Optimizely has a free one, as does Evan Miller's online calculator) and input these variables. The output is the number of unique visitors you need in each variant before calling a result.\n\nFor a page with 3% baseline conversion and a 20% MDE, you typically need around 8,000 to 12,000 visitors per variant. At 1,000 visitors per week, that test needs to run for 16 to 24 weeks to reach statistical significance. If your traffic is too low to run a clean test in a reasonable timeframe, do not run the test. Implement the change as a best-bet improvement and measure the before-and-after instead.\n\n---\n\n## Quick Wins: Changes That Rarely Need Testing\n\nSome changes have enough evidence behind them that you can implement them without running a formal test, especially if your traffic volume is too low to support reliable testing.\n\nAdd specificity to the headline. \"Project management for teams\" becomes \"Project management for engineering teams of 20 to 200.\" Specificity increases relevance for the right audience.\n\nRemove phone number field from the lead form. Phone number fields consistently reduce form completion rates by 5 to 20% when they are not necessary for the follow-up process. If your team does not call prospects within 24 hours, remove the field.\n\nMove the primary CTA above the fold. If your main CTA button requires scrolling on most devices, move it up.\n\nAdd a \"no credit card required\" statement near the trial CTA. This single phrase regularly increases trial sign-ups by 5 to 15%.\n\nReplace a generic testimonial with a specific, quantified one. \"Great product!\" does no conversion work. \"We reduced our reporting time from 8 hours per week to 45 minutes\" does.\n\nAdd a comparison table to the pricing page. Buyers who are deciding between your tiers need to know what they gain by upgrading. A visible table with specific feature differences makes that decision easier.\n\nReduce form fields to three or fewer for top-of-funnel offers. Every additional field reduces completion rate. Ask only for what you need to deliver on the offer.\n\n---\n\n## Common CRO Mistakes\n\nTesting without enough traffic: A test that reaches significance at 70% confidence is not reliable. It will produce false positives more often than you want. Wait for 95% confidence.\n\nStopping tests early: Peeking at results and stopping a test when it looks like the variant is winning is called \"optional stopping.\" It dramatically inflates false positive rates. Set your endpoint in advance and stick to it.\n\nRunning too many tests simultaneously: Tests on the same page at the same time contaminate each other unless they are properly isolated (for example, using mutually exclusive audience segments). Run one test on each major conversion point at a time.\n\nIgnoring seasonality: A test that runs from January to February produces different results than the same test run in Q3, because visitor intent and composition change by season. Account for this when interpreting results.\n\nCalling a test based on relative lift alone without checking absolute numbers: A 50% relative improvement from 0.4% to 0.6% conversion is probably within the noise. Look at absolute conversion rate and total conversions, not just relative lift.\n\nNot learning from failed tests: A test that shows no improvement or a decline is not a failure. It is information. Document what you tested, what you expected, what happened, and what you concluded. The best CRO programs are built on a library of this evidence.\n\n---\n\n## Output Formats\n\nFor a CRO audit, deliver:\n\nA friction inventory: a list of every friction point identified, organized by category (copy, design, form, trust, offer), with the evidence behind each one.\n\nAn ICE-scored prioritized opportunity list: the top ten opportunities with scores and rationale.\n\nThree to five fully written hypotheses: in the format specified above, covering the highest-priority opportunities.\n\nSample size calculations for each test: to help the user assess which tests are feasible given their traffic volume.\n\nQuick wins list: changes to implement immediately without waiting for tests.\n\nFor a single A\u002FB test design, deliver:\n\nThe hypothesis, the control description, the variant description, the primary metric, the secondary metrics, the estimated sample size needed, and the recommended test duration based on current traffic.\n",{"name":5,"source":24,"license":15,"description":25,"imported_at":26,"source_path":27,"source_repository":28},"https:\u002F\u002Fgithub.com\u002Fscayver\u002Fmarketing-skills","Use this skill for conversion rate optimization on marketing pages and forms. Trigger phrases include: \"improve our conversion rate,\" \"CRO audit,\" \"why isn't our landing page converting,\" \"analyze our homepage conversions,\" \"optimize our pricing page,\" \"run a CRO analysis,\" \"write a hypothesis for an A\u002FB test,\" \"what should we A\u002FB test,\" \"design an A\u002FB test,\" \"calculate sample size for a test,\" \"find friction on our page,\" \"quick CRO wins,\" or \"optimize our lead capture form.\" This skill applies to homepages, landing pages, pricing pages, and lead capture forms. Distinct from signup flow optimization and in-app paywalls.\n","2026-09-20","skills\u002Fcro\u002FSKILL.md","scayver\u002Fmarketing-skills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fadvertising\u002Fcro","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill cro",{"tool":41,"command":42},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill cro",{"tool":44,"command":45},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill cro","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fadvertising\u002Fcro\u002FSKILL.md",true,[],[50,58,65,72,79,86],{"skillNo":51,"name":52,"nameZh":6,"description":53,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":57},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":59,"name":60,"nameZh":6,"description":61,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":57},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":57},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":57},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":57},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":57},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]