Marketing analytics prototype
Ad Performance Analysis Agent
A campaign-performance analysis prototype that parses ad-data CSVs, compares weekly metrics, highlights risks and wins, and uses Gemini to help produce actionable summaries.
Product work and architecture overview by Savi Saluwadana.
What it does
- Who it serves
- Marketing operators and teams reviewing campaign performance.
- Product purpose
- An advertising analytics and reporting prototype.
- CSV parsing of campaign dates, spend, clicks, impressions and conversions.
- Week-over-week metrics and threshold-based risk/win flags.
- Gemini-assisted summaries and email reporting paths.
How it works
- An operator supplies campaign data as CSV.
- The analysis module normalizes rows and compares current and previous periods.
- Threshold logic identifies risks and wins; Gemini helps summarize next steps.
- The reporting interface displays results and the email path can send a summary when configured.
Architecture
React / Vite → Express → CSV analysis + Gemini → email
| Component | Responsibility |
|---|---|
| React interface | Campaign-data input and analysis views. |
| Express API | Prototype configuration, scheduling and reporting endpoints. |
| Analysis module | CSV normalization, weekly comparison and threshold flags. |
| Gemini + Resend | Configured AI summarization and email-report delivery. |
Current scope
Google Ads and Supermetrics connection endpoints are currently mocks. Scheduled demonstration runs use generated sample CSVs and in-memory configuration, rather than a durable live account-ingestion pipeline.
Architecture summary reviewed . The repository is private; this page provides a public product summary.