Ecom Heads
I had Claude build me an ads analysis tool. It took less than an hour and I use it every week.
Last week I uploaded a search terms report to Claude. It told me exactly what to negate, what to promote to exact match, and what needed more data. Formatted and ready to paste into Google Ads.
Useful. But it was a one-off. Then I realized I could save that entire setup as a reusable tool inside Claude Projects. Now I open it every week, upload the new report, and get the same analysis without rebuilding anything.
Not a prompt. A program.
🤝 Connect with John on LinkedIn
Say hey, I'll respond!
Connect
The Gap
The search term review you keep skipping? Claude does it in 90 seconds.
Every operator knows this work matters. Almost none do it consistently.
The Work That Never Gets Done
The conversation around AI in advertising is dominated by platform automation: Advantage+, pMax, Smart Bidding, automated creative. All of it hands control to the platform.
Claude is the opposite play. It gives you leverage on the manual analysis work that almost never happens. The stuff that should happen weekly but gets skipped because it takes too long.
Search term reviews · Creative performance audits · Campaign post-mortems
Competitive ad analysis · Budget reallocation modeling
Every operator knows these matter. Almost none do them consistently. Because 45 minutes of manual work doesn't happen when you have three other fires burning.
Claude does it in 4 minutes. Build it right, and you do it by uploading one file.
Three Levels: Pick Where You Start
Level 1: Just use the prompt (no setup, works today)
Export your search terms report as a CSV. Open Claude. Paste it with:
"Here's my Google Ads search terms report for the last 30 days. I sell [product]. Categorize these into: terms to add as exact match keywords, terms to add as negatives, and terms that need more data. Format the negatives as a comma-separated list ready to paste into Google Ads."
You'll get structured output in under a minute. Works for Meta creative reports too: paste ad performance data, ask Claude to identify what's fatiguing and what angles are winning.
Level 2: Build a saved tool in Claude Projects (Opus 4.6)
Claude Projects let you save instructions, context, and logic so you come back without rebuilding. I built a search terms analysis tool in a Claude Project using Opus 4.6. I gave it my product context, my negative keyword strategy, my match type preferences, and examples of past decisions I'd made on borderline terms.
Now I open the project, upload this week's report, and it runs the same analysis every time. Not a prompt I reconstruct. A tool that exists and waits for my data.
Start with the prompt from Level 1, then ask Claude to turn it into a saved project with the right context baked in.
Level 3: Connect directly to the ad platforms (no exports)
Claude can write code to pull data from the Google Ads API or Meta Marketing API directly. At that point the analysis runs automatically. No exports, no uploads.
For Google Ads specifically, there's an even simpler path:
Google Ads Scripts. JavaScript snippets that run inside your account on a schedule. Claude writes them from plain English:
"Write a Google Ads script that emails me whenever a campaign spends more than 80% of its daily budget before 3pm."
Paste the output into Tools > Scripts in your Google Ads account. Set the schedule. Done.
12 Things You Can Actually Build
These are real workflows, not hypotheticals.
1
Search term triage (Google/Bing): Upload your report weekly. Flags negatives, surfaces new keyword opportunities, formats for direct import. I run this every Monday. 90 seconds per account.
2
Meta creative health report: Export ad-level data for the last 14 days. Claude identifies fatiguing creatives, winning angles, and suggests what to test next. One operator I talked to caught a creative fatigue issue that would've burned $2K+ before he noticed manually.
3
Budget pacing alert (Google Ads Script): Runs hourly, emails you when spend is tracking ahead or behind your daily target. Essential for managing 5+ campaigns.
4
ROAS drop detector: Compares last 7 days vs. prior 7 days per campaign. Alerts you to drops above a threshold you define. Replaces daily dashboard checking.
5
Campaign post-mortem tool: Paste a performance timeline alongside the changes you made. Claude correlates your actions with the data. I used this after a rough two weeks on a client account and it caught a negative keyword blocking a high-converting long-tail term in minutes.
6
Competitive ad monitor: Pull new ads from the Meta Ad Library for a list of competitors. Claude analyzes angles, offers, hooks, formats. Weekly competitive intelligence without paying for a monitoring tool.
7
Negative keyword conflict checker: Upload your negative keyword lists and active keyword lists. Claude cross-references them and flags conflicts. Happens more often than anyone admits, especially on accounts with years of accumulated negatives.
8
Landing page vs. ad copy alignment audit: Paste your top 5 ad headlines/descriptions alongside landing page copy. Claude scores message match and flags disconnects. One of the most common reasons for high CTR + low conversion rate.
9
Klaviyo flow performance review: Export flow analytics. Claude identifies declining open/click rates, drop-off points in sequences, and where to test new subject lines. 2 minutes from any ESP.
10
Google Ads quality score audit: Export keyword report with quality score columns. Claude groups by tier, identifies which keywords are dragging down your account, and recommends specific ad copy or landing page changes. Most operators check this quarterly at best.
11
pMax asset group analyzer: Export pMax asset performance. Claude categorizes "Best" vs. "Low" assets and identifies patterns. Useful because Google's native pMax asset reporting is, being generous, not great.
12
Weekly spend vs. forecast tracker: Paste monthly budget targets and actual weekly spend. Claude builds a pacing model and tells you whether to increase or decrease daily budgets to hit your monthly number. Runs every Monday, eliminates the "we overspent by 20%" conversation.
The Practical Point
Most operators running $10K-$200K/month in ad spend skip analysis work that would directly improve performance. Search term reviews, creative audits, budget pacing checks: these are pattern recognition tasks. Claude is fast at pattern recognition.
The gap isn't capability. Nobody has taken an afternoon to build the tool.
Start with Level 1 this week
Pick the analysis task you skip most often. Paste your last 30 days of data into Claude with a clear prompt. If the output is useful, spend an hour building it into a Claude Project with Opus 4.6 so it's reusable. That's the version you actually use every week.
Talk soon,
John Sciacchitano
Ecom Heads: Scale or Die Trying