Performance Ad Production AX Automation
I connected Claude Code and Notion so competitor research, weekly marketing analysis, ad planning, and production prompts flow into live performance-ad creative.
Role & contribution
- Role
- Product Business Lead · AX workflow design and implementation
- My contribution
- Designed and implemented the automation and prompts using Claude. Selected prompts and generated ad images with Gemini and GPT.
- Collaboration
- Performance marketer: campaign application and weekly performance dashboards.
- Scope
- Market and competitor research → Ad briefs → Image generation · Six months live · Six briefs per week.
Context
When competitor research, marketing analysis, creative planning, and image production operated separately, the same handoffs and decisions had to be repeated every week.
I redesigned the workflow as one data flow where analytical evidence becomes a direct input to planning and production.
Claude Code automation routines
On a weekly schedule, the system reviews competitor advertising and market signals, creates a marketing analysis, and records it in Notion.
It then reads the analysis and generates six ad briefs with production-ready image prompts.
Weekly marketing ad analysis
Weekly competitor research and ad analysis accumulate in a Notion database so prior observations and new hypotheses remain traceable.
Analysis and ad briefs live in the same operating database, allowing production results to become an input to the next week's strategy.
Ad planning and creative output
I designed and implemented the automation and prompts using Claude, then selected prompts and generated ad images with Gemini and GPT.
The performance marketer applied the creatives to live campaigns and prepared weekly performance dashboards. We used these results to inform the next briefs.
- Competitor research and analysis
- Weekly
- Ad briefs and production prompts
- Notion
- Ad briefs
- 6 per week
- Live operation
- 6 months
Campaign performance analysis
Creative produced through the automated workflow was deployed in FB_UA_iOS and FB_UA_AOS.
In the latest actual week, the two campaigns generated 260 installs with a combined CPI of ₩1,491 and D1 ROAS of 10.2%.
iOS AAA delivered a ₩1,204 CPI, 24.8% below its ₩1,600 target, demonstrating stronger acquisition efficiency. AOS AAA reached 18% D1 ROAS, 3.6× its 5% target, demonstrating stronger revenue efficiency. These differences informed an OS-specific direction for the next creative and campaign cycle.
- Combined installs
- 260
- Combined CPI
- ₩1,491
- Combined D1 ROAS
- 10.2%
- AOS ROAS vs. target
- 3.6×