Docs · Usage

What to ask your behavioral data

Once your AI is connected over MCP, these are the capability areas it can query — each with a prompt worth stealing.

Capability areas and example prompts

Store overview & opportunities

A CRO overview of the whole store — where sessions go, where money leaks, sized opportunities ranked by revenue impact.

Try
"Give me a CRO overview of the last 30 days. Where am I losing the most revenue, and what's the single highest-impact fix?"

Conversion funnels

Step-by-step funnel analysis from product view to checkout completion, weighted by cart value.

Try
"Walk my funnel for the last two weeks. Which step loses the most cart value, and how does mobile compare to desktop?"

Click & heatmap analysis

What gets clicked, ignored, hovered, and rage-clicked — per page and per element, with revenue context.

Try
"On my best-selling product page, which elements do buyers click that non-buyers ignore?"

Cart analysis

What enters carts, what gets abandoned, and what abandoned carts had in common.

Try
"Analyze abandoned carts from last week. What products, cart values, and behaviors show up most?"

Revenue correlation

Which behaviors correlate with buying — the evidence layer for UX decisions.

Try
"Which on-page behaviors correlate most strongly with completed orders this month? What should I change first?"

Session journeys

How sessions actually move through the store — paths, patterns, exits, and summaries of representative sessions.

Try
"Summarize the typical journey of sessions that exit on the shipping-cost step. What did they do before leaving?"

Forms, search & site health

Form field friction, on-site search behavior, JS errors, and page performance — the technical leaks.

Try
"Are any JS errors or slow pages correlated with funnel drop-offs this week?"

Before/after comparison

Honest comparison of a metric across two windows — for verifying whether a change worked.

Try
"I changed my PDP layout on the 10th. Compare add-to-cart rate and revenue per session before and after, and tell me if the sample is big enough to trust."

Three habits that make answers better

  • Give a time range."Last 30 days" or "compare this week to last" produces sharper tool calls than an open-ended question.
  • Ask for the money, not the clicks. Frame questions in revenue terms — the data model carries cart and order context precisely so answers can be revenue-weighted.
  • Demand statistical honesty.Add "tell me if the sample is too small to trust" to comparison questions. A good assistant will flag noise instead of narrating it.

Frequently asked

Do I need to know the tool names to ask questions?

No. You ask in plain language — 'where does my checkout leak?' — and the AI discovers the available MCP tools and picks the right ones. The capability areas on this page exist so you know what kinds of questions have data behind them.

How much history do the tools see?

Whatever your plan's captured sessions cover — there's no artificial lookback cap on MCP queries. Ask for 'the last 30 days' or 'compare this week to last week' and the tools take a date range.

Will the AI make up numbers?

The design goal is the opposite: answers are grounded in tool results, and good assistants cite the data they pulled. If a sample is too small to support a conclusion, a well-prompted assistant will say so — and you should ask it to. Statistical honesty matters most on low-traffic stores.

Can I automate recurring questions?

Yes — anything you can ask once you can put on a schedule with your AI tooling, like a Monday-morning funnel report or an anomaly check. Claude Code and similar clients support scheduled or scripted MCP workflows.

Steal a prompt and ask it today.

Every prompt on this page runs against your store's real behavioral data over MCP. Free plan up to 5,000 sessions/month; paid plans from $39/month.

Install on ShopifySee how it works