A Practitioner’s Dissertation
How to Build a World-Class GA4 Setup
A Practitioner’s Dissertation, Grounded in a Production Implementation
Abstract
Most Google Analytics 4 installations record data without measuring anything a business can act on. This dissertation documents the audit and reconstruction of one such property from first principles, and generalizes the work into a method any practitioner can follow.
The subject property had passed every casual inspection. It fired a rich, well-named event stream, defined custom dimensions, and marked several conversions. It also hid a class of silent failures that no dashboard surfaces. The server-side conversion event posted to https://www.google.com/mp/collect, a host that returns HTTP 404, so no server-recorded lead ever reached GA4 across the property’s lifetime. A generic click event sat marked as a conversion, ready to poison bidding the moment the account linked to Google Ads. Event data retention sat at the two-month floor, deleting the seasonal history a demand-seasonal business depends on. The one signal engineered for resilience did not exist.
The reconstruction proceeded in a fixed order that the dissertation argues is the correct one: correctness before architecture, architecture before activation, activation before optimization. Along the way it produced hard numbers. Server-side leads attributed for the 69 percent of submissions that carry a first-party analytics cookie, and skipped, rather than faked, for the remaining 31 percent. A single Admin API call reversed the retention loss. A multi-agent adversarial audit of nine measurement pillars surfaced findings that a single reviewer missed, and corrected several plausible claims that did not survive fact-checking.
The dissertation contributes four things: a measurement maturity model that grades a property on correctness rather than feature count; a correctness-first implementation method; an offline-revenue architecture that closes the loop between a form fill and a booked job for a lead-generation business; and a governance and privacy model sized for a small operator rather than an enterprise. The thesis throughout: world-class measurement means a correct, closed revenue loop and an honest account of its limits, not a maximal feature set.
A Note on Anonymization
The client’s brand identity is anonymized so the work can be shared; everything technical is preserved verbatim, because the technical substance is the point. Throughout the paper the company name reads {COMPANY}, its production domain reads {company-domain}, and the company name inside identifiers reads {company}. The GA4 identifiers, every event and parameter name, the industry (pest control), and the region (four Central California counties) are preserved.