I recently came across a brilliant workflow shared by a performance marketing expert on LinkedIn that completely changes how we look at ad optimization. If you run paid traffic, you know the drill: staring at spreadsheets, trying to guess why one headline worked and another flopped, all while burning cash on testing.
This industry pro shared a method that essentially replaces the need for a junior media buyer or a hefty agency retainer using nothing but Claude and a CSV file. I was genuinely impressed by how this creator structured the data analysis to force the AI into thinking like a strategist rather than just a copywriter.
The best part? The author points out that this setup takes about 15 minutes and costs nothing if you already have the tool. Here is the detailed breakdown of how this solo marketer operates like a full team.
The Setup Phase
The original poster emphasizes that you shouldn’t just dump this into a standard chat window. The real power comes from using the “Projects” feature, which allows for long-term context retention.
- Navigate to Claude.ai
Start by logging into your account. The expert notes this works best on the desktop interface where you can manage files easily. - Access the Projects tab
Look for “Projects” on the left sidebar. This is distinct from your historical chat logs. - Initialize your workspace
Click to create a new project and name it “Ad Copy.” This creates a dedicated container for all your brand assets.
The Data Phase
This is where the magic happens. The creator explains that AI is only as good as the data you feed it. Instead of asking for generic ideas, you are going to provide hard performance metrics.
- Open your ad platform
Go to Google Ads or Meta Ads Manager (or whichever platform you use). - Export your campaign data
The author specifies that you need to export your data as a CSV file. Crucially, ensure your export includes headlines, impressions, clicks, and conversions. The AI needs to see the relationship between what was said (the headline) and what the user did (the click/conversion). - Upload to Claude
Take that CSV file and upload it directly into your “Ad Copy” project.
The Context Phase
I think this is the most critical part of the expert’s workflow. Most people skip this and get generic results. The author insists on “priming” the project with brand guardrails.
- Add brand instructions
In the Project instructions area, define your brand. The creator suggests including your tone of voice, target audience details, and significantly, a list of “words you never use.” This prevents the AI from sounding like a robot. - Feed it winners
Don’t just show it what failed; show it what worked. The expert recommends adding past headlines that actually converted and your best Calls to Action (CTAs). This gives Claude a benchmark for success.
The Execution Phase
Once the data and context are loaded, the innovator behind this post provided a specific prompt to trigger the analysis. This prompt is designed to turn the AI into a “performance marketing strategist.”
- Paste this exact prompt
The author crafted this prompt to generate a table that is ready for immediate implementation:
“You’re my performance marketing strategist. I’ve uploaded my ad data with performance metrics.
→ Identify my 10 lowest-performing ads by CTR.
→ Write one sentence on why it’s underperforming.
→ Generate 5 new headline options per ad.
→ Generate 3 new description options per ad.
→ Write benefit, urgency & curiosity-driven angles.
→ Flag any patterns across my worst performers I should stop repeating.
Format this as a table to copy into a spreadsheet.”
The Output
According to the expert, running this prompt generates a massive amount of actionable data instantly. You aren’t just getting text; you are getting a diagnostic report.
- 50+ Fresh Headlines: These aren’t random; they are iterated versions of your ads based on your specific performance data.
- 30+ Descriptions: Ready-to-test variations that match your brand voice.
- Failure Analysis: A clear picture of exactly why certain ads are failing (e.g., “too vague,” “lack of urgency”).
- Pattern Recognition: The AI flags bad habits you might not catch when staring at an Excel sheet, such as overusing certain buzzwords or passive voice.
Why This Matters
The creator makes a compelling point about the long-term value of this setup. This isn’t a one-time trick. Because you are using the Projects feature, Claude “remembers everything.”
Next week, when you have fresh data, you simply upload the new CSV. The AI already knows your brand voice, it knows what worked last time, and it knows what mistakes to stop suggesting. As the author puts it, this is how “one person operates like a full team.” You are essentially building a custom ad agency that gets smarter every time you use it.
It’s a powerful reminder that the best way to use AI isn’t just for writing, but for analyzing the gap between effort and results.
Check out the full post from this savvy professional to see the discussion around this technique.