Field notes on automated tracking QA, tag governance and analytics testing — written by the consultants who ran these audits by hand before automating them.
Pixel Helper is green, Test Events light up, and Events Manager counts 1,340 purchases on a day the back office closed at 780. The three levels of verification, and the one that decides how your media budget gets spent.
The tag is green, the reports show traffic, and revenue is 17% below the back office. The three levels of verification, and the one everybody skips.
The real risk of a replatform is not that tracking breaks, it is losing the ability to say whether the drop came from the site or the collection. The three deliverables that make the question decidable.
The document everyone cites and nobody maintains. What it must contain, how to build one, and the four mistakes that make it useless.
One regression every two weeks, and two weeks before anyone notices — except on revenue, which gets fixed in hours. What 8 years of dataLayer audits look like in numbers.
A green warehouse, dbt tests passing, and a revenue figure that still doesn't match the back office. The three layers data quality lives on, and the one almost nobody monitors.
ObservePoint builds one Journey per event, on CSS or JS selectors — what holds up at compliance scale becomes a full-time load at the scale of a full tracking plan. The real cost of that maintenance, and where MayIA° fills the gap.
GA4 drowns in tutorials, Piano Analytics has almost none. The method we run on client projects: the tools that actually help, the 8 checkpoints in order, and the mistakes that come back project after project.
Opening the hood and understanding what you're looking at: the event name, its properties, and what the SDK adds by itself. With a real production hit taken apart line by line.
A hit leaving before anyone clicks the banner isn't always a fault — in advanced consent mode it's expected. How to tell the two apart in the Network tab, and fix the order when it really is broken.
TrackingPlan is excellent at alerting you when production breaks. It was never built to stop it from breaking in the first place. The real boundary between the two, and how to tell which side your problem is on.
The foundation of all tracking, explained simply: what a data layer holds, how to structure it, how to set it up with GTM — and why the one that works today can break without a sound on the next release.
The items array, the funnel events in order, a complete end-to-end purchase, and the pitfalls that quietly report the wrong revenue without telling anyone.
Double-counted page_view, purchase with no amount, empty items, inconsistent casing… Seven classic errors, each with its symptom, cause and fix. Not one of them raises an alert.
When to test, what to check, in what order. The step-by-step method, the control-point checklist, and where doing it by hand runs out when you ship every week.
A perfect data layer guarantees nothing: between the push and the report, a hit can be blocked, truncated or duplicated. Where it breaks, and how to verify it really arrives in GA4, Meta and Segment.
Three tools, three answers to the same problem — and one decision axis nobody looks at: who actually writes the test logic, and who keeps it current as your taxonomy evolves.