‹ 首页

ab-testing

@ominou5 · 收录于 1 周前

A/B testing strategy and implementation for funnel pages. Defines what to test, how to structure variants, statistical significance thresholds, and common testing patterns.

场景转化优化漏斗测试实验设计

适合你，如果你需要科学地测试和优化营销漏斗页面的转化效果。

/ 下载安装
ab-testing.skill双击,或拖进 Claude 桌面版 / Cowork,即完成安装↓ .skill↓ .zip
用别的 agent?下载 .zip 解压,把文件夹放进它的技能目录
Claude Code~/.claude/skills/(项目级 .claude/skills/)
Codex CLI~/.codex/skills/
Cursor自动读取上面两处目录
其他工具见其文档的「skills」目录;两个下载是同一份文件,只是名字不同
/ 通过 npx 安装 校验哈希
npx oh-my-skill add ominou5/funnel-architect-plugin/ab-testing
/ 通过 bash 安装
curl -fsSL https://oh-my-skill.com/install.sh | bash -s -- ominou5/funnel-architect-plugin/ab-testing
/ 已经装过?验证本机副本,不用重装
npx oh-my-skill verify ominou5/funnel-architect-plugin/ab-testing
安装目标可用 --agent / --scope 或 --to 明确指定;省略时只会在唯一已存在的 agent 目录上自动选择,零命中或多命中会停止并提示。content_hash 缺失或不一致均拒装。
67GitHub stars
~611上下文体积 · 单文件
镜像托管

怎么用

技能原文 SKILL.md作者撰写 · MIT · 7c5acf6

A/B Testing

Test everything. Opinions are nice — data is better.

What to Test (Priority Order)

| Priority | Element | Expected Impact | |---|---|---| | 🔴 P0 | Headline | 10–50% lift | | 🔴 P0 | CTA text + color | 5–30% lift | | 🟡 P1 | Hero image/video | 5–20% lift | | 🟡 P1 | Form fields (fewer vs. more) | 10–40% lift | | 🟡 P1 | Social proof placement | 5–15% lift | | 🟢 P2 | Page layout (long vs. short) | 5–20% lift | | 🟢 P2 | Pricing display | 5–25% lift | | 🟢 P2 | Urgency messaging | 3–15% lift | | 🔵 P3 | Color scheme | 2–10% lift | | 🔵 P3 | Font choices | 1–5% lift |

Testing Rules
  1. Test one variable at a time — Change only the element being tested
  2. 50/50 split — Equal traffic to each variant
  3. Minimum sample size — At least 100 conversions per variant before calling a winner
  4. Statistical significance — Wait for 95% confidence before declaring a winner
  5. Run for at least 7 days — Captures day-of-week variations
  6. Document everything — Record hypothesis, variant details, and results
Test Hypothesis Template
HYPOTHESIS: If we change [element] from [current] to [proposed],
then [metric] will [increase/decrease] by [estimated %]
because [reasoning based on conversion principles].

TEST SETUP:
- Control (A): [Current version description]
- Variant (B): [New version description]
- Primary metric: [Conversion rate / Click rate / etc.]
- Secondary metric: [Revenue / Engagement / etc.]
- Required sample: [Number] visitors per variant
- Estimated duration: [X] days at [Y] daily visitors
Common Tests by Page Type
Opt-In Page
  • Headline: Problem-focused vs. Solution-focused
  • CTA: "Get Free Access" vs. "Download Now" vs. "Send Me the Guide"
  • Form: Email only vs. Name + Email
  • Social proof: Subscriber count vs. Testimonial
Sales Page
  • Long-form vs. Short-form copy
  • Video sales letter vs. Text
  • Testimonials at top vs. After offer
  • Payment: One-time vs. Payment plan (default)
Pricing Page
  • 2 plans vs. 3 plans
  • Annual default vs. Monthly default
  • Feature comparison table vs. Simple list
  • "Most Popular" badge placement
Results Tracking

After each test, log:

TEST: [Test Name]
DATE: [Start] → [End]
TRAFFIC: [Total visitors] ([Per variant])
RESULTS:
  Control: [X]% conversion ([N] conversions)
  Variant: [Y]% conversion ([N] conversions)
WINNER: [Control/Variant]
LIFT: [+/- X]%
CONFIDENCE: [X]%
NEXT: [What to test next based on learnings]
按 MIT 许可原样转载,未经改动 · 在 GitHub 查看 →

评论

登录即可评论;带「已验证安装」的,是发布者名下有本店的安装或持有记录。
…