No BI SaaS, no monthly rent, no "hang on, let me pull the report." This is the exact architecture running in production for our brands — and why we prefer our own code over Looker.
In nearly every ecommerce brand we've audited, the same thing happens: Meta says one thing, Shopify another, Google another, and someone on the team loses Monday morning copying cells into a sheet to "make it tie out" — if they're lucky. The decision that matters (scale, pause, change creative?) gets made on Tuesday, with Friday's data.
The usual answer is to rent a BI tool (Looker Studio, Power BI, Triple Whale). Our problem with that: the connectors fail exactly where it counts — the fine reconciliation between platforms — and they add a monthly rent for life to a problem you can solve with three scripts.
The whole system is three pieces you already know:
The secrets (API tokens) live in GitHub Secrets — never in the code. The repo is private. And because the deploy fires from the Action, the dashboard updates even when everyone's laptop is off.
A dashboard that "roughly ties out" is worse than no dashboard: nobody trusts it and everyone goes back to the sheet. That's why every run checks that what Meta reports as attributed revenue and what Shopify records as paid sales close to the cent — and if the difference (Δ) drifts out of range, it fires a push alert instead of publishing broken numbers.
API access to your platforms (read-only), a private GitHub repo and a Vercel account. That's it. If your team doesn't code, this is literally what we build under the Data & Dashboards capability — dashboard running from week 1.
Three weeks, read-only access, no cost, and a written scope you keep, whether you continue or not.
NEXT ON THE BLOG → YOUR REAL ROAS DOESN'T LIVE IN META: IT LIVES IN THE SHOPIFY RECONCILIATION