AgentConversion

Conversion funnel audit

Sessions to checkout, stage by stage — with the biggest leak ranked by $ impact and the source-level cut from GA4.

Works with
ClaudeChatGPTAny MCP-compatible AI tool
Requires
StoreHero MCPShopifyGA4
Get the prompt
Example chat in Claude. You ask: "Run the Conversion Funnel Audit on the last 30 days". Claude reads your StoreHero data and answers: The biggest leak is product page to cart: add-to-cart fell from 18.8% to 17.1%, worth about $6,300 in contribution margin. It adds an interactive example chart, Funnel, sessions to orders, with example data you can switch below.
Funnel, sessions to ordersExample

Example, Last 30 days: 2.8% conversion, down from 3.1%

  • Sessions84,200
  • Product views46,300
  • Add to cart7,900
  • Checkout4,100
  • Orders2,350

Sessions: Sessions are up 6% on the prior 30 days.

The prompt

How to run it

Paste it into a new chat in Claude or ChatGPT and you get the answer straight away.

Use the StoreHero MCP to do the following analysis. Always anchor the answer in real StoreHero data, not platform-reported metrics.

Run a full conversion funnel audit for the last 30 days vs. the prior 30 days. Pull the website analytics page in StoreHero alongside the store overview to see funnel and revenue together. The goal is to understand where users are dropping off, why, and which fixes would most impact contribution margin — not just conversion rate in isolation.

Build the funnel across these stages and quantify drop-off at each step:

1. Total sessions
2. Bounce rate
3. Add-to-cart rate (% of sessions)
4. Checkout-initiated rate (% of sessions)
5. Conversion rate (% of sessions)

For each stage, compare current vs. prior period and flag whether the change is volume-driven or rate-driven. Critical: don't read a falling conversion rate as a problem in isolation. If sessions are down but AOV, gross profit per order, and contribution margin are up, the store may be filtering for higher-intent traffic — which is healthier, not worse. State this explicitly when you see it.

Then layer in source-level conversion rate from GA4 (organic search, paid search, paid social, direct, email, referral). Identify which sources are pulling the blended conversion rate up and which are dragging it down. Cross-reference paid sources against StoreHero blended CPA to identify whether low-converting paid traffic is also unprofitable, or just lower-converting at acceptable cost.

End with three sections: (1) the single biggest funnel leak ranked by $ impact, (2) what likely caused it (sessions quality shift, on-site friction, checkout issue, product/pricing perception), (3) two deeper dives worth running next — typically the "Product page conversion analysis" and "Landing page audit" tiles.

At the bottom of your answer, render an inline visual summary: a vertical funnel chart showing all 5 stages with absolute numbers and conversion % at each step, plus KPI cards for blended conversion rate, AOV, gross profit per order, and contribution margin (each with vs prior period and traffic-light status). Add a small horizontal bar showing source-level conversion rate from GA4. Place text commentary alongside the visuals. Use the Spend Advisor tool to suggest what to explore next.

Get the full prompt

Enter your email to see every prompt in full and copy it in one click. It unlocks the whole library.

See our privacy policy.

How to use a skill

New to skills? It takes three steps and a few minutes: connect StoreHero once, copy a skill into Claude or ChatGPT, and ask.

  1. Connect StoreHero to Claude or ChatGPTAdd the StoreHero MCP connector once. It works with Claude, ChatGPT and any other tool that supports MCP. Setup guide
  2. Pick a prompt and copy it into ClaudeOpen one, press Copy prompt (or download the file) and paste it into a new chat.
  3. AskIt pulls your live StoreHero data and gives you the analysis in minutes. Ask follow-up questions in the same chat.

Run every prompt on your own numbers.

Connect your store to StoreHero and the whole library works on your live profit data.