This week is about one trade most teams get backwards. They cut the work that makes money and keep the work that just fills a calendar. Today we go deep on analytics, the part of performance marketing that tells you which is which. By the end you will know how to point a model at your own numbers and let it find the busywork hiding in plain sight.

Step one. Do not guess what is working. Open Google Analytics 4, go to the admin panel, and create a service account key under data access. That key lets a machine read your traffic without you clicking through reports. You now have a door into your own numbers that a model can walk through.

Step two. Take that key and feed it to Claude Code through an MCP server for GA4. An MCP server is a small adapter that lets the model call a tool directly. Now the model can ask your analytics questions in plain words and get real rows back, instead of you exporting a spreadsheet at midnight.

Step three. Ask the model one blunt question. Which landing pages bring people who buy, and which ones only bring people who leave. The model runs the query, sorts by conversion, and hands you a short list. The pages at the bottom are your busywork. You have been feeding them for no return.

Step four. Most teams track page views and call it a day. Set up real conversion events instead, a form sent, a call booked, a cart paid. In GA4 you mark these as key events. Now the model is counting money moments, not mouse moves, and the whole picture stops lying to you.

Step five. Ask the model to break revenue down by source. Not clicks, revenue. You will often find one quiet channel doing the heavy lifting while a loud one eats the budget. The model can show you this in seconds because it is reading the same data, just without the wishful thinking.

Step six. Look at the dashboards you build every week. Ask honestly who opens them. Most go straight to a folder. Have the model write one short weekly summary in plain language instead, pulled live from GA4. One paragraph people read beats ten charts nobody does.

Step seven. Set a simple rule. Each morning the model pulls yesterday's numbers and flags anything that fell hard. No human staring at a screen. You only get pinged when something actually breaks, which means you stop checking and start fixing only when it counts.

Here is the honest part. We do not have a secret dashboard we hide from clients. The whole studio runs on this same loop, the model reading our own analytics and telling us where we waste effort. When we say cut busywork and keep revenue, we mean we did it to ourselves first.

So that is the takeaway. Analytics is not about more charts. It is about pointing a model at your real numbers so it can name the work that pays and the work that does not. The next step is simple. Go see the whole thing running in public at office dot temerarii dot xyz.