A tool that audits Google Ads and Meta Ads exports, normalizes heterogeneous data and turns it into analytical signals before AI interpretation.
The problem
Advertising exports use different naming conventions, structures and numeric formats. Without an audit layer, conclusions can be drawn from data that is not comparable.
The solution
I separated facts from reasoning: parsing, normalization and calculations stay deterministic; the model receives structured context to interpret relevant signals.
What I built
I built a React and Vite application with a Node and Express backend to import CSV files, detect platforms and available fields, remove aggregate rows and calculate KPIs and account benchmarks.
How it works
CSV exports ? data audit ? field taxonomy ? normalized metrics ? benchmarks ? signals ? AI interpretation.
Engineering choices
Field taxonomy and localized number parsing are deterministic. OpenAI explains limitations and signal priorities instead of replacing calculations.
Outcome
A repeatable flow from fragmented exports to output closer to an analyst workflow.
Gallery
