The Ecomm Analyst

Yes, and not by uploading a CSV. Several analytics platforms now publish a connector that lets Claude or ChatGPT query your live commerce data directly. You authorize it once, then you ask questions in plain language and the assistant pulls real numbers from your account rather than guessing.

The underlying standard is called MCP, which is just a protocol for letting an assistant call a defined set of tools. You do not need to know anything about it. In practice you are adding a connector in your assistant’s settings and signing in.

You are not connecting Shopify directly

This is the part that trips people up. Most of these connectors do not point at Shopify. They point at an analytics platform that has already ingested your Shopify orders alongside your Meta spend, your Google spend, your Klaviyo sends, and whatever else you have connected. The assistant queries that layer, not your store.

That distinction matters because it determines what you can ask. If the analytics platform never ingested your subscription data, the assistant cannot answer subscription questions no matter how well you phrase them. The connector inherits whatever the platform underneath it knows, and nothing more. Before you evaluate the assistant experience, look at what the platform actually holds.

Who publishes one

Polar Analytics documents an official connector for both Claude and ChatGPT, running on a semantic layer that unifies its commerce and ad sources into consistent metric definitions. Triple Whale documents one as well, connected either through a sign-in flow or an API key depending on which assistant you use, and it is not included on the free plan.

ThoughtMetric, which sponsors this blog, publishes a connector that exposes its metric and dimension catalog to the assistant, so you can ask for something like new-customer ROAS grouped by campaign and get it back without opening the dashboard. Northbeam is the interesting exception. There are community-built servers floating around for it, but I could not find an official one documented on Northbeam’s own site. If you are on Northbeam and this matters to you, ask their team directly rather than trusting a third-party listing.

That gap is worth noting generally. Directory sites list plenty of connectors that vendors did not build and do not support. A community server means somebody wrote code against a public API. It may work fine. It also may break silently when the API changes, and nobody is on the hook to fix it.

What the setup actually looks like

In Claude you open settings, find connectors, and either pick the vendor from the directory or add a custom one with a server URL. In ChatGPT it is a similar flow under connectors. Either way you sign in with your existing account for that platform and approve access. Most of these require a paid tier on the assistant side, so a free Claude or ChatGPT account will not do it.

Total time is a few minutes if your analytics platform is already set up. If it is not, that is the real project, and the connector is the last five percent of it. I wrote up the step-by-step for one of these in this walkthrough on connecting ThoughtMetric to Claude.

What it is good at, and what it is not

It is very good at the questions you would otherwise not bother asking. Pull spend by campaign for the last six weeks and tell me which ones moved. Compare new-customer ROAS across channels this month against last. These are questions where the answer is worth thirty seconds but not worth ten minutes of clicking, so in practice you skip them. Removing the friction means you ask them.

It is worse at anything requiring judgment about what the numbers mean. The assistant will happily tell you a channel’s ROAS went up without telling you that you changed the attribution window last Tuesday. It does not know your business context unless you give it. Treat the output as a fast query result, not as analysis.

The other thing I would flag is that the assistant can misread a result and state it confidently. I have seen one grab the wrong date range and present the answer with no hedging at all. Spot check anything you plan to act on against the dashboard. That habit costs you very little and will eventually save you from a bad decision.

Is it worth setting up

If you already pay for a platform that publishes one, yes. It costs you a few minutes and changes how often you interrogate your own numbers, which is the actual benefit. If you do not have that platform yet, the connector is not a reason to buy one. Buy the platform because the underlying measurement is right for you. The assistant access is a convenience layer on top of a decision you should make on other grounds.