Why I’m starting to run ads
Most of Cohoon Consulting’s clients so far have come from word of mouth, referrals, and a slowly growing blog. That’s the local-first playbook working as intended. But there’s a missing piece: I’ve never actually paid for traffic to my own site.
If I’m going to advise clients on Facebook ads — which is increasingly the conversation — I should be running them myself with my own money on the line. Skin in the game. So that’s the experiment.
What an MCP server actually is (in plain English)
MCP stands for Model Context Protocol. The short version: it’s a way to give Cursor (the tool I build everything in) the ability to talk to other systems — Facebook’s Ads API, Google Analytics, Google Business Profile, whatever — without leaving the editor.
The longer version, for fellow nerds: an MCP server runs locally on my machine, exposes a set of tools (“list ad accounts,” “create campaign,” “adjust budget”), and Cursor can call them through a clean conversational interface. Authentication happens once. After that, the tool is just there, like any other command.
The practical version: I can say “show me how my ad performed yesterday” in plain English and Cursor pulls the numbers. No tab-switching to Ads Manager. No clicking through six menus.
The workflow change
Before MCP, running a Facebook ad meant:
- Switch to the browser, open Ads Manager.
- Find the right account, click through to the right campaign.
- Adjust the thing I came to adjust.
- Go to a different tab to check GA4.
- Try to remember what I was doing in Cursor before all that.
With MCP, it’s:
- Open Cursor.
- “What did yesterday’s spend look like, and which ad got the most clicks?”
- “Pause the worst-performing ad.”
- Back to the thing I was building.
The whole loop is in one window. No context switch. That’s the real unlock.
Honest numbers (week 1)
Spend: small to start, intentionally. I’m treating week one as a learning budget, not a scaling budget. Numbers below are real and I’ll keep updating this section as the data comes in.
Week 1 setup
One campaign, two ad sets (Oakland County and Genesee County), three creative variations each. Daily budget intentionally conservative while I watch what’s working. Target: local trades and service business owners.
Conversion goal
Visit to /contact.html, tracked via GA4. The goal isn’t a click — it’s a real inquiry. I don’t care if 200 people see the ad and bounce.
I’ll add the click-through rate, cost per click, and inquiry counts here as the test runs. The whole point of build-in-public is that the numbers stay honest, including when they’re bad.
What I’m watching next
- Cost per inquiry, not cost per click. A click is vanity. An actual contact form submission is the only number that matters.
- Which creative pulls. I’m running a build-in-public photo, a comparison-style screenshot, and a plain-text-on-color variant. Whichever wins teaches me what my market responds to.
- Whether the MCP workflow actually saves time over Ads Manager. If it doesn’t, I’ll go back to the browser. Tools are servants, not statements.
- What happens when I push the budget up. Cheap clicks at $5/day rarely scale to cheap clicks at $50/day. I’ll know in a few weeks.
One more thing worth knowing
Running ads through an MCP server isn’t a magic shortcut. The campaign still has to be good. The targeting still has to be right. The landing page still has to convert. All the boring fundamentals of Facebook ads still apply.
What MCP unlocks is removing the friction between “I have a question about my ads” and “I have the answer.” That alone is enough to change how often I actually look at the numbers — which is the part most local advertisers get wrong.
If you’re a local Oakland or Genesee County business owner thinking about running Facebook ads — or you’re curious how I’m using Cursor and MCP servers to run an entire business out of one window — book a free conversation and I’ll walk you through the whole stack.