What was built
A Worker that, every 30 minutes, cross-referenced ad revenue with sessions and conversions, calculated the value per page, and sent that value back to Google Ads as a conversion value — so bidding would optimize for return, not clicks. Revenue dashboards, marketing automation integration, and sync to BigQuery.
What was learned — and it’s worth more than the code
- Audit directly against the APIs, never against processed endpoints or dashboards: I found attribution discrepancies that only showed up when comparing the raw sources.
- Google API rate limits, quotas, and retry rules are product-specific; ignore them and you get silently throttled.
- The contract ended with a negative return for the client, and that’s on record: the engine worked, the arbitrage model’s economics didn’t. I’d rather show the case than hide it.
What it’s useful for today
Companies that depend on Google Ads/GA4 and don’t trust their own numbers: reconciliation across sources, a pipeline to BigQuery, and alerts for what doesn’t add up.