Six tabs, six versions of the truth
Monday morning. Shopify open in one tab. Klaviyo in another. Meta Ads Manager, Google Ads, GA4, and that spreadsheet you've been patching together since 2022.
Every one of those tools gives you its own answer. Meta says its campaigns drove the week. Klaviyo says email did. Google says it was search. Add up everything each platform claims credit for and you'll find you've apparently sold far more than your bank account agrees with.
In the founder calls I have each week, this comes up again and again. Plenty of data. Very little insight. And the decision that gets made by default is the one the loudest dashboard suggests: spend more on ads.
This is where MCPs come in. They're the quiet bit of plumbing that's changing how I work with every brand I'm involved with, and I think they're the most important thing an ecommerce marketer can get their head around right now.
So what is an MCP?
MCP stands for Model Context Protocol. It's an open standard that lets an AI assistant like Claude or ChatGPT plug directly into the tools you already use and read the data inside them.
Think of it as a universal plug. Each tool builds one MCP server. Each AI assistant knows how to talk to any MCP server. Connect Shopify, Klaviyo and your analytics, and you can ask questions of all of them at once, in plain English, without exporting a single CSV.
Anthropic introduced it, and in December 2025 donated it to the Linux Foundation's new Agentic AI Foundation, co-founded with OpenAI and Block. That matters for you. It means this is a shared industry standard you can build your way of working on, rather than one vendor's feature that might vanish next year.
There are two directions to it, and most of the noise this year has been about the wrong one for most founders:
Shopper-facing. AI shopping agents querying your store. If you're on a paid Shopify plan, your store already has a live MCP endpoint you never had to switch on. Good catalogue data is the job here.
Operator-facing. You connecting your own stack to an AI assistant so you can interrogate your business. This is where the real leverage sits for a £1m to £3m brand, and it's what the rest of this article is about.
The vendors have moved fast. Shopify, Klaviyo, Google Ads, GA4 and Meta Ads all now run official MCP servers, and Klaviyo expanded its integration across Claude in May 2026. Most merchants simply haven't noticed yet.
The insight lives between the tools
Here's the bit that gets me excited. Every tool in your stack is very good at answering its own questions. Klaviyo knows how your flows perform. Meta knows what a click cost. Shopify knows what sold.
The questions that actually move a business sit across those tools, in the gaps nobody's dashboard covers:
Which acquisition channel brings in customers who come back for a second order?
Which first-purchase product leads to the highest twelve-month value?
When customers tell you how they found you, does it match what your ad platforms are claiming?
Is your post-purchase email treating a likely repeat buyer the same as a one-and-done bargain hunter?
Until recently, answering any of those meant an analyst, a data warehouse or a very long weekend with a spreadsheet. With your stack connected through MCPs, it's a conversation. You ask, the assistant pulls from each source, and you get a joined-up answer in minutes.
That's what I mean by collective data. Your tools stop being separate scoreboards and become one shared backdrop of insight you work from every day. It's the same principle I keep coming back to: optimisation built on the data you collect, the technology you master and the experience you deliver. MCPs make the first two far easier, so you can spend your energy on the third.
An optimisation route: "our CAC keeps rising"
Let's make it real. A founder tells me Meta costs are climbing and their agency's answer is more budget. Here's the route I'd take with a connected stack, one question at a time:
Start with unit economics. Ask Lifetimely which first-purchase products lead to a second order within 90 days, and what those customers are worth after a year.
Hear it from the customer. Ask Fairing how those repeat buyers said they found you. Compare it with what Meta and Google are claiming.
Check the experience after the sale. Ask Klaviyo how the post-purchase flow performs for that group against everyone else.
Bring it together. Ask the assistant to lay all three side by side and tell you where they disagree.
The kind of thing that surfaces is rarely dramatic. A product that looks average on day-one ROAS but quietly brings in your best long-term customers. A channel your survey responses credit far more than your attribution does. A post-purchase flow sending the same three emails to everyone.
None of those answers says "spend more". They say shift creative towards the product that builds repeat customers, give the under-credited channel a fair test, and branch the flow so likely repeat buyers get a different journey. Small, specific changes that compound. That's optimisation, and it's the work that turns the repeat-purchase lever from a nice idea into something you can measure.
Where to start
You don't need every tool connected on day one. A lean stack covers most of what a £1m to £3m brand needs:
Shopify for orders, products and customers.
Klaviyo for flows, campaigns and segments. Its MCP is hosted, so connecting it is a sign-in rather than a developer job.
A post-purchase survey tool such as Fairing, so the customer's own voice sits alongside the attribution numbers.
A unit economics layer such as Lifetimely, for cohorts, repeat rate and real profit.
A data connector such as Windsor.ai to pull Meta, Google Ads and GA4 into the same conversation.
Then ask the questions you've always wanted answered and never had the time to dig into. A few to try this week:
What's our repeat purchase rate by first product bought?
Which customers bought once in the last six months and are now overdue a second order?
Where do survey responses and ad attribution disagree most?
Which Klaviyo flow drives the most second orders, and which drives the least?
Write the good questions down. Ask them again every month. That habit is where the compounding starts.
Three things to watch
Messy data stays messy. Connecting your tools won't fix inconsistent product names, broken UTMs or a returns process that never reaches Shopify. The assistant will join up whatever you give it, confidently. Spend an afternoon on data hygiene first and every answer after that gets better.
Read before you write. Many MCP servers can change things as well as read them: create segments, edit flows, update products. Start with read-only access. Get comfortable asking questions before you let anything act on your behalf.
The answer is a backdrop, the decision is yours. MCPs give you a sharper picture of what's happening. They don't know your brand, your customer's quirks or why that one product matters to the story you're telling. Use the insight to decide what to work on, then bring the human touch: the positioning, the copy, the reason someone hops off the fence and buys from you.
So, are they the key?
I think they are. The job of an ecommerce marketer has always been to understand the customer better than anyone else and act on it. For years the data needed to do that sat in separate boxes, and only the big brands could afford to join them up.
MCPs hand that ability to a lean team, or a founder with an hour on a Monday morning. Before you put another pound into Meta, connect what you already have, ask better questions, and let the answers set the agenda for the week.
Next time, I'll walk through the exact questions I run across a connected stack in the first week with a new client.
Ian
Founder of EcommerceGrowth.com