Sales Funnel Analysis & Optimization Skill
One command maps your site's full visitor-to-purchase path, pinpoints drop-off stages, and prices every fix in revenue terms.
The skill is prompt-only: it fetches a target site, writes FUNNEL-ANALYSIS.md, and has no hard dependencies (requirements.txt lists only optional reportlab). No covert exfiltration, credential access, or destructive defaults found, so the risk surface is small. Deducted for: no user confirmation or least-privilege boundary declared for fetching arbitrary URLs, no data-flow disclosure (page content sent to the model), no rollback/removal mechanism, and no guidance on robots/anti-scraping respect. Score 12.
Internally self-consistent: phases, scoring frameworks, and output template align; cross-skill file references are explicit. Deducted for: static review cannot reproduce key paths (site fetching, funnel step discovery, conversion estimation); no failure handling or diagnosable error design for unreachable sites, login walls, JS-rendered pages, or SPAs; metrics like 'estimated load time' are guesses that may produce precise-looking but fabricated numbers. Capped by static calibration; score 9.
Clear scenario (funnel analysis of any website), broad funnel-type taxonomy and benchmark tables, and a semantically precise trigger command. Deducted for: no declared non-fit boundaries (unreachable sites, non-English sites, logged-in funnel stages), no guidance when real data is unavailable beyond generic benchmarks, no Chinese-language support, and no assessment of reachability of overseas sites from mainland China. Score 7.
Well-layered documentation (phases → frameworks → output template), README with install/uninstall, clear MIT license. Deducted for: no versioning, changelog, known-limitation disclosure, or maintenance ownership in the SKILL.md; hidden assumptions about suite-level script dependencies; placeholder rather than real sample outputs. Score 8.
Substantial value as a structured framework: funnel-type identification, per-page scoring, RPV methodology, and prioritization matrix directly guide optimization, with marginal value over ad-hoc analysis. Deducted for: conversion figures lean on generic benchmarks rather than measured site data; revenue-lift estimates ('expected lift 10-30%') are unverified heuristics; output usability depends heavily on in-the-moment model judgment, unverifiable statically. Score 6.
Auditable primary material exists (full methodology, scoring frameworks, benchmark tables), but no third-party execution evidence, no tests, no CI covering this skill; benchmarks are uncited, 'expected lift' ranges have no sources, and fact is not separated from inference. No conclusion independently reproducible from a static read; score 3.
- Conversion and revenue-lift figures in outputs are largely generic benchmark estimates, not measured site data; verify manually before presenting to clients.
- The skill fetches arbitrary user-provided URLs; confirm the target site permits fetching, and note page content will be sent to the model for processing.
- For login-walled, JS-rendered, or sites unreachable from mainland China, analysis may be incomplete or fail, with no explicit failure-feedback design.
- Publisher is unverified; the repository lacks versioning and changelog, and long-term maintenance status is unknown.
- This assessment is a purely static source review; the skill was not executed and confidence is low.
What does this skill do, and when should you use it?
This is the /market funnel sub-skill inside the AI Marketing Suite for Claude Code. Given a URL, it detects the funnel type (eight templates from SaaS trial to e-commerce to webinar), traces each visitor step, scores every page on clarity, continuity, motivation, friction, and trust, and benchmarks conversion rates and revenue per visitor against industry tables. The output is a FUNNEL-ANALYSIS.md report with prioritized recommendations and revenue-impact estimates. It is a pure instruction skill: all methodology lives in SKILL.md, with no dedicated scripts of its own.
Fetches the target URL and walks each funnel page; classifies the funnel type; builds structured step records including exit points, friction, and trust elements; draws an ASCII funnel map with estimated per-step conversion rates; scores pages on a five-dimension framework; computes metrics like Visitor→Lead and Trial→Paid plus RPV and LTV:CAC against benchmark tables; produces P1–P5 prioritized fixes with expected lift ranges (headline A/B tests, pricing redesign, form-field reduction, risk reversal, etc.); audits pricing pages and lead magnets against checklists; maps funnel stages to email sequences; and writes the full analysis to FUNNEL-ANALYSIS.md.
- A SaaS founder preparing a pricing-page redesign wants to know which fields and missing trust signals cause trial-to-paid drop-off
- An agency runs the analysis before a client pitch and uses the revenue-opportunity figures to justify its services
- An e-commerce operator with below-benchmark cart-to-purchase conversion needs a prioritized checklist for checkout fixes
- A solopreneur wants to see where their free-trial funnel sits against the 0.5–3% visitor-to-customer benchmark
- A webinar host wants to check whether registration, attendance, and sale rates hit benchmarks like 40–55%
What are this skill's strengths and limitations?
- Complete methodology: funnel-type detection, five-dimension page scoring, and RPV-based revenue quantification form a reusable framework
- Built-in benchmark tables (e.g., Trial→Paid 10–15% good, Cart→Purchase 50–60% good) make findings comparable rather than arbitrary
- P1–P5 priority matrix weighs both expected lift and implementation effort, producing directly schedulable output
- Cross-integrates with the suite's audit, copy, email, and competitor reports, fitting a broader marketing workflow
- Without real analytics data, conversion figures rely on estimation and benchmarks rather than instrumentation; accuracy depends on estimate quality
- Pure Markdown instructions with no verification scripts; page fetching is subject to anti-bot and JS-rendering limits the source doesn't address
- Revenue-impact figures (e.g., '10–30% lift') are generic experience ranges, not site-specific predictions
- Requires the whole-suite installer; standalone installation of this sub-skill is undocumented
How do you install this skill?
Install the full suite (this skill ships inside it): curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-marketing-claude/main/install.sh | bash, or clone the repo and run ./install.sh. MIT licensed; uninstall with ./uninstall.sh. Optionally pip install reportlab for PDF reports (a suite-level feature, not required by this sub-skill). The source does not document installing this sub-skill on its own.
How do you use this skill?
Run /market funnel <url> in Claude Code. The skill fetches the site, analyzes the funnel, and saves FUNNEL-ANALYSIS.md. If MARKETING-AUDIT.md, COPY-SUGGESTIONS.md, EMAIL-SEQUENCES.md, or COMPETITOR-REPORT.md already exist, it cross-references them; suggested follow-ups are /market copy, /market emails, or /market landing for deeper work.
How does this skill compare with similar options?
Within the suite, /market landing (landing-page CRO deep dive) and /market audit (full-site audit including a conversion dimension) overlap partially; this skill focuses specifically on the end-to-end funnel path and revenue quantification. The source names no external alternatives to compare against.