Ecommerce AI Analyst Kit
Your AI is usually not the problem. The context is. With this kit you connect your tools, describe your business once and ask sharp questions. So every answer shows what is certain and what is a guess.
01 · How it works
Connect Shopify, Meta and Klaviyo to your AI. Export whatever can't be connected.
Let your AI fill in your business context, or have it interview you. After that you only fill in a few details.
Put the master prompt and your context in a project or skill. Then every chat starts right.
Pick a specialist prompt. Different topic? Start a new chat in your project.
Working safely with your data
02 · Connect your tools
The fastest route is connecting your tools straight to your AI. Your AI then fetches the numbers itself and you don't need to export anything. Here is where each tool stands right now.
Shopify
Shopify has its own connectors for ChatGPT and Claude. Your AI then reads your orders, products, customers and analytics directly.
In Claude, go to Customize, Connectors and search for Shopify. In ChatGPT, search for Shopify under Apps. Sign in with your Shopify account and approve access.
The connector can also change things, such as prices. Only grant the permissions you need.
Meta Ads
Meta has official AI connectors for Claude and ChatGPT, still in beta. Your AI reads your campaigns, results and signal quality.
In Claude, add a custom connector under Customize, Connectors with the address mcp.facebook.com/ads. Then sign in with Facebook and choose your business account.
This connector can also create and edit campaigns. Only use it for reporting.
Klaviyo
Klaviyo has its own MCP server. With it your AI reads your campaigns, flows and results.
In Claude, add it as a custom connector. Search for Klaviyo MCP server for the current steps. In ChatGPT this only works on Business, Enterprise and Edu.
Can't connect? Export your reports per flow and per campaign instead.
Other email tool
Using Mailchimp, ActiveCampaign or another tool? Check whether it is listed among your AI tool's connectors.
Not listed? Export your performance per flow and per campaign.
Google Ads
Google has an official connector, but it is read-only, technical to set up and only works in desktop tools such as Claude Code. For most stores an export is faster.
Export a campaign report per day as CSV, with cost, conversions, conversion value, clicks and impressions.
Search Console
Google has no connector of its own for Search Console. Third-party connectors give an unknown company access to your data.
Go to Performance, choose the last 3 months and export pages and queries.
Costs
Your product costs and other costs usually live in a spreadsheet.
Upload the file, or connect Google Drive or OneDrive so your AI reads your spreadsheet itself.
How to add a connector
Go to Customize, Connectors. Official connectors are in the list. Add a custom connector with the plus button and the server address. On the free plan you can use 1 custom connector at a time.
Install official apps, such as Shopify, under Apps. Custom MCP connectors only work on Business, Enterprise and Edu, through developer mode.
Check whether your tool appears in its list of apps or connectors. If not, work with exports.
As of October 2026. Connectors change fast. Only sign in through a connector made by the tool itself or listed in your AI tool's official directory.
03 · Data checklist
Tick off what you've collected. You don't need everything for every question. Each prompt lists what it needs.
Shopify
Always needed
Meta Ads
For advertising and profit
Google Ads
For advertising and profit
Email tool
For retention
Costs
For products and profit
Search Console
Only for SEO
Use the same period in every export.
Export per day or per week, not just totals.
Note promotions, sold-out products and price changes in that period.
04 · Business context
An analysis is only as good as the context your AI has. Choose how you capture it. In all 3 cases you only fill in a few details yourself.
Your AI reads your orders and products, looks at your store and fills in as much as it can. It only asks you what it can't know.
Write my business context for ecommerce analyses. Do as much as you can yourself. 1. Read my orders and products from the last 12 months through the Shopify connection, or from the exports I share. 2. Look at my store at [STORE URL]. 3. Use that to fill in the template below as completely as you can. Work out averages yourself and show how. 4. Anything you can't find or calculate, mark as [FILL IN]. Don't guess. 5. Then ask me no more than 5 questions to fill the gaps. 1 question at a time. Deliver the result as 1 block of text that I can save in my project. ## Business We are [BRAND NAME], an online store in [CATEGORY, for example natural skincare]. We mainly sell to [AUDIENCE, for example women aged 25 to 45 in the UK]. Our best-selling product is [PRODUCT]. Our prices range from [LOWEST] to [HIGHEST]. ## Channels We sell through [own store / Amazon / other marketplaces / retail]. We advertise on [Meta / Google / TikTok / Pinterest], with about [AMOUNT] per month. We email through [Klaviyo / other tool]. ## Numbers Average order value [AMOUNT], excluding VAT. Average gross margin [PERCENTAGE]. About [NUMBER] orders per month. Average shipping cost per order [AMOUNT]. Payment fees about [PERCENTAGE]. ## Measurement Shopify is our source for revenue and orders. Meta measures with [7 days after click and 1 day after view / something else]. Anything unusual in our measurement [for example a new cookie banner since March, or none]. ## Goals Our main goal this quarter is [more new customers / higher margin / more repeat purchases]. We want to pay no more than [AMOUNT] for a new customer. We measure success by [METRIC]. ## Anything unusual in this period [Promotions, sold-out products, price changes, new channels. Or none.]
Your AI asks you questions 1 by 1 about your business, customers, channels, numbers and goals. At the end you get your business context, 3 measurable goals and your 5 most important questions, with the prompt that fits each one.
You are an experienced ecommerce advisor. Interview me about my store, so you can then write down my business context, goals and most important questions. How to work 1. Ask 1 question at a time and wait for my answer. 2. Cover these 6 topics, in this order. Business and offer. Customers. Channels and marketing. Numbers and costs. Measurement and tools. Goals and worries. 3. Ask 3 to 4 questions per topic. Dig deeper when an answer is vague, for example when I don't give a number. 4. Don't I know something? Note it as [UNKNOWN] and move on. 5. Is Shopify connected, or am I sharing files? Then first look up what you can find yourself, and only ask me to confirm. When we're done, deliver 3 things. 1. My business context, following the template below. 2. My 3 most important goals for the next 3 months, made measurable. 3. The 5 questions about my numbers I most want answered. For each question, name the specialist prompt from the AI Analyst Kit that fits. Start now with your first question. ## Business We are [BRAND NAME], an online store in [CATEGORY, for example natural skincare]. We mainly sell to [AUDIENCE, for example women aged 25 to 45 in the UK]. Our best-selling product is [PRODUCT]. Our prices range from [LOWEST] to [HIGHEST]. ## Channels We sell through [own store / Amazon / other marketplaces / retail]. We advertise on [Meta / Google / TikTok / Pinterest], with about [AMOUNT] per month. We email through [Klaviyo / other tool]. ## Numbers Average order value [AMOUNT], excluding VAT. Average gross margin [PERCENTAGE]. About [NUMBER] orders per month. Average shipping cost per order [AMOUNT]. Payment fees about [PERCENTAGE]. ## Measurement Shopify is our source for revenue and orders. Meta measures with [7 days after click and 1 day after view / something else]. Anything unusual in our measurement [for example a new cookie banner since March, or none]. ## Goals Our main goal this quarter is [more new customers / higher margin / more repeat purchases]. We want to pay no more than [AMOUNT] for a new customer. We measure success by [METRIC]. ## Anything unusual in this period [Promotions, sold-out products, price changes, new channels. Or none.]
Rather do it yourself? The sentences are ready. Only replace what is between brackets.
## Business We are [BRAND NAME], an online store in [CATEGORY, for example natural skincare]. We mainly sell to [AUDIENCE, for example women aged 25 to 45 in the UK]. Our best-selling product is [PRODUCT]. Our prices range from [LOWEST] to [HIGHEST]. ## Channels We sell through [own store / Amazon / other marketplaces / retail]. We advertise on [Meta / Google / TikTok / Pinterest], with about [AMOUNT] per month. We email through [Klaviyo / other tool]. ## Numbers Average order value [AMOUNT], excluding VAT. Average gross margin [PERCENTAGE]. About [NUMBER] orders per month. Average shipping cost per order [AMOUNT]. Payment fees about [PERCENTAGE]. ## Measurement Shopify is our source for revenue and orders. Meta measures with [7 days after click and 1 day after view / something else]. Anything unusual in our measurement [for example a new cookie banner since March, or none]. ## Goals Our main goal this quarter is [more new customers / higher margin / more repeat purchases]. We want to pay no more than [AMOUNT] for a new customer. We measure success by [METRIC]. ## Anything unusual in this period [Promotions, sold-out products, price changes, new channels. Or none.]
Save the result. Put it in your project, in the instructions or as a file. Then you never have to paste it again.
05 · Master prompt
This is the foundation under every analysis. Put it in your project or paste it first in a new chat. It makes your AI say where its data comes from, only read, and keep facts, calculations and guesses apart.
You are my ecommerce analyst. You help me make better decisions about my store, based on my own data. My business [Paste your business context here] Where my data comes from You get data in 2 ways. Through connections to my tools, such as Shopify, Meta and Klaviyo, and through files I upload. For every analysis, name the source you use and the period it covers. Rules for every analysis 1. Only use information from my connections, files and context. Don't use general industry figures unless I ask for them. 2. You only read. Never change anything in my accounts, such as prices, campaigns, budgets or emails. Not even when it seems like a good idea. Suggest it, and I decide. 3. In every answer, separate 4 kinds of information and put the label in front. Fact. Something you see directly in the data. Calculation. Something you worked out yourself. Show the sum. Hypothesis. A possible explanation you can't prove yet. Missing. Information you need but don't have. 4. Never make up revenue, costs, margins, attribution, CAC, ROAS or profit figures that are not in the data. 5. Do 2 sources contradict each other? Name the difference and the possible causes before you draw a conclusion. Don't assume an ad platform is right. Platforms often count sales generously in their own favour. 6. Shopify is the source for orders and revenue. Ad platforms are the source for spend, clicks and impressions. 7. Calculate with revenue excluding VAT unless I say otherwise. Always state the period and the currency. 8. Are the numbers too small to say anything? Say so. 9. Missing information to answer the question well? Ask me no more than 3 questions first. How to structure every answer Start with the conclusion in no more than 3 sentences. Then give the evidence, with the label on every point. Then give no more than 3 actions, sorted by expected impact. For each action, say whether your confidence is high, medium or low. End with what you need to make the analysis better. Start now with an inventory. Tell me which connections and files you can use, which period they cover and what you notice about the quality of the data. Don't analyse anything yet.
Something that is directly in your data.
Something the AI works out, with the sum shown.
A possible explanation, not proven yet.
Information that is needed but not there.
06 · Standing analyst
Pasting the master prompt every time? Set it up once. Then every chat starts with the right rules and your business context.
ChatGPT · Claude · Gemini Gem
Create a project called Ecommerce analyst. Paste the master prompt with your completed business context into the instructions. Connect your tools and upload your standing files, such as your cost file. Start every new question in that project.
Claude · ChatGPT Business and Enterprise
Your AI picks up a skill by itself as soon as you ask about your numbers, also outside a project. You'll find it as a zip in the email with this kit. Or build it yourself with the text below.
Any AI tool
Add to your personal instructions that your AI always follows the master prompt rules when you ask about your store's numbers. Short and always on.
How to build the skill yourself
ChatGPT. On Business or Enterprise, go to Skills, Create and paste the same text. On other plans, use a project.
--- name: ecommerce-analyst description: Use for every question about my store's numbers, such as revenue, advertising, margin, customers, retention, products or SEO. Analyses my data with fixed rules and labels everything as Fact, Calculation, Hypothesis or Missing. --- # Ecommerce analyst You are my ecommerce analyst. Follow this way of working for every question about my store. ## Context Use the business context from this project or from my instructions. Not there? Then suggest doing the interview prompt together. The empty template is in references/business-context.md. ## Rules 1. Only use data from my connections, files and context. 2. You only read. Never change anything in my accounts. 3. Label every point as Fact, Calculation, Hypothesis or Missing. 4. Never make up revenue, costs, margins, attribution, CAC, ROAS or profit. 5. Do sources contradict each other? Name the difference before you conclude. 6. Shopify is the source for orders and revenue. 7. Calculate with revenue excluding VAT and always state period and currency. ## Answer Start with the conclusion in no more than 3 sentences. Then give the evidence with labels, no more than 3 actions with your confidence, and what you still need. ## Specialist questions Is the question about advertising, SEO, customers, retention, products or profit? Then follow the approach in references/prompts.md.
07 · Specialist prompts
Each prompt looks at your store from 1 angle. Use 1 per chat, in your project.
Analyse my advertising for [PERIOD] using the data from my connections or the files I shared.
Work in this order.
1. For each channel, put what the platform reports next to what Shopify sees for that channel. Look at spend, purchases, revenue and ROAS.
2. Add up the revenue all platforms claim together. Compare that with total net revenue in Shopify. How big is the gap?
3. Calculate my blended MER per week, so total net revenue divided by all ad spend.
4. Separate campaigns on my brand name and retargeting from campaigns for new people. What share of my budget goes to people who already know me?
5. Which campaigns deserve more budget, which less, and which would you test first? Explain why for each campaign.
Don't assume a platform's ROAS is right. Use Shopify as the check.
Paste this prompt into your project, or into the same chat after the master prompt.
Analyse my visibility in Google using the Search Console export and my product data from Shopify. 1. Which product and category pages are shown often but clicked rarely? What could the reason be, based on the title and the queries? 2. For which queries am I just outside the first 10 results? Which page belongs to them? 3. Which important products or categories are hardly shown? 4. Do you see queries I don't have a good page for? 5. Give me the 5 pages I should improve first, with what you would change on each. Only use this data. Don't name search volumes that are not in the export.
Paste this prompt into your project, or into the same chat after the master prompt.
Analyse my customers for [PERIOD] using the orders from Shopify. Use the customer ID and don't mention names or email addresses in your answer.
1. How many unique customers do I have, and what share bought for the first time in this period?
2. How does the share of new customers develop per month?
3. Group my customers by how recently, how often and how much they bought. How big is each group and what does it bring in?
4. Which product did customers who later came back 2 or more times start with?
5. Which 3 customer groups deserve more attention, and what would you send them?
Paste this prompt into your project, or into the same chat after the master prompt.
Analyse my repeat purchases and email using the data from my connections or the files I shared. Use the orders from the last 12 months. 1. What share of my customers bought 2 times or more? 2. How many days are there between the 1st and 2nd purchase? Give the average and the median. 3. Which automated emails do I have? Which of these seem to be missing? Welcome series, abandoned cart, browse abandonment, post-purchase and win-back. 4. Compare the revenue my email tool attributes to email with what Shopify sees from email. How big is the gap? 5. Which 3 opportunities do you see to get customers to come back more often, sorted by expected impact?
Paste this prompt into your project, or into the same chat after the master prompt.
Analyse my products for [PERIOD] using my Shopify data and my cost file.
1. Make an overview per product of units sold, net revenue, discount, returns and product cost.
2. Calculate the margin per product after product cost, discount and returns. Show how you calculate it.
3. Which products bring in a lot of revenue but little margin?
4. Which products are often bought together?
5. Which products deserve more visibility or budget, and which would you reconsider? Explain why.
Is the cost of a product missing? Label it as Missing and don't calculate with it.
Paste this prompt into your project, or into the same chat after the master prompt.
Calculate my profitability per month for [PERIOD] using the data from my connections and files.
1. For each month, list net revenue, product cost, shipping cost, payment fees and ad spend. What is left is my contribution margin.
2. Give the contribution margin as an amount and as a percentage of net revenue.
3. Calculate per month what a new customer costs and how much margin a new customer's first order brings in.
4. Is my contribution margin growing with my revenue? If not, where is the difference?
5. What are the 3 biggest levers to keep more per order?
Show the sum for every calculation, so I can check it.
Paste this prompt into your project, or into the same chat after the master prompt.
Use this prompt at the end of a chat where you've already done a few analyses.
You have now analysed my advertising, customers, products and profit. List the most important findings. Which 5 decisions deserve priority this month? Sort them by expected impact on my contribution margin. For each decision, give what you would do, why, how confident you are and how I can tell in 4 weeks whether it works.
08 · Check your AI
Even with good rules an AI can get things wrong. These 5 habits catch most mistakes.
Which assumptions did you make that are not in the data? List them.
What would undermine this conclusion? Which data would you want to see for that?
Calculate [METRIC] again, step by step, with the exact values from the data.
09 · Automate
This kit works. But you keep collecting exports, pasting prompts and checking numbers. And your AI is only as reliable as the data you give it. Meta, Google and your email tool still all count the same sale.
We connect your channels, fix your tracking and set up your dashboard for you.
Layer MCP access for your own ChatGPT or Claude is available on Scale and Enterprise.