Data QA
Verify that yesterday's and last week's conversions actually landed: tags firing, deduplication holding, import lag accounted for. Anomalies get flagged before any other work starts.
This page is the manual we actually work from, published in full. It exists for the same reason our reports carry source annotations: a claim that cannot be traced does not count. Below is what happens to an account in a normal week, in what order, and why.
All tags live in Google Tag Manager and fire from data layer events. No hardcoded gtag snippets in templates: everything is versioned, auditable and reversible inside the container.
Each conversion fires once. Transaction and lead IDs deduplicate across pages, tags and reloads, so a refreshed thank-you page never doubles the count.
Where the booking system or CRM allows it, completed appointments and sales are uploaded back to the ad platforms, so bidding optimizes on outcomes rather than form fills.
One fixed naming convention for source, medium and campaign. No free-text UTMs, so reports aggregate cleanly instead of splitting the same campaign across five spellings.
What was spent, against what was budgeted, with pacing for the rest of the month.
Week-over-week movement on the metrics that matter, each annotated with the source it was read from.
Every bid, budget, negative and creative change from the week, with the reasoning behind it.


The next planned steps, so nothing in the account moves without being announced first.
