Reverse a trend of rising CPAs by using LLMs to analyze historical ad performance data and identify non-obvious patterns.
Results
CPA Reduction: -25%
Conversion Rate: +18%
Analysis Time: Days -> Hours
Methodology
Extract
Analyse
Hypothesise
Test
Iterate
Strategy & Execution
Data Analysis: Extracted and anonymized dataset from Meta Ads Manager. Fed data into ChatGPT to identify correlation patterns between ad copy and audience segments.
Hypothesis Testing: Developed structured A/B tests based on AI findings.
Iterative Loop: Established a feedback system where test results refined the AI's future recommendations.
What I Delivered
Anonymised Dataset: Extracted and cleaned Meta Ads historical performance data ready for AI-assisted analysis.
AI Analysis Report: ChatGPT-powered correlation mapping of ad copy variables vs. audience segment performance.
A/B Test Framework: Structured hypothesis-to-test workflow built from AI findings and run across live campaigns.
Optimisation Playbook: Iterative feedback loop where test results continuously refined AI recommendations each cycle.
Performance Outcome: CPA reduced by 25%, conversion rate up 18%, analysis time compressed from days to hours.