Machina
← How to Build a World-Class GA4 SetupAppendix G

Appendix G·Appendix

References and Further Reading

A practitioner dissertation asserts how a product behaves, so it owes the reader the product's own documentation for each behavioral claim. This appendix collects the primary sources behind the technical assertions in the preceding chapters. Every entry is a Google first-party document: an API reference, a developer guide, a product help article, or an official developer blog post. Third-party tutorials and commentary are deliberately excluded, because the value of a reference here is that it is the authority the product's own team maintains, not a secondary reading of it.

Each entry carries a one-line note naming the specific claim it supports and the chapter that makes that claim. The reader who doubts a number or a mechanism in the body can go straight from the assertion to the source that stands behind it.

Two caveats govern the whole appendix. First, every URL was checked in July 2026, and Google's documentation moves: sections get renamed, help-center answer IDs get merged, and developer guides get restructured on a schedule Google does not announce. A dead link a year from now is a documentation move, not a retracted fact; search the document title. Second, product behavior changes faster than this paper can. The offline-conversion upload path is the clearest example in this very set, where a mechanism that was standard in early 2026 was foreclosed for new accounts by mid-2026. Read each source as true as of the field work, and re-verify anything load-bearing before you rely on it in a new build.

G.1 The Measurement Protocol

The server-side event path is the spine of the reconstruction, and its single most consequential bug was a wrong hostname. These references define the correct endpoint, the required payload fields, and the validation surface.

  • Measurement Protocol (Google Analytics 4) reference. https://developers.google.com/analytics/devguides/collection/protocol/ga4/reference Supports Chapter 6 and Appendix B. Defines the collection endpoint on the google-analytics.com host and the required client_id plus events array. The pre-reconstruction code posted to www.google.com/mp/collect, a host this reference never names, which is why every server lead returned HTTP 404 and never reached the property.

  • Sending Measurement Protocol events to Google Analytics. https://developers.google.com/analytics/devguides/collection/protocol/ga4/sending-events Supports Chapter 6 and Chapter 9. Documents client_id, session_id, and engagement_time_msec, the fields that decide whether a server event joins the right session and appears in Realtime. The chapter's insistence on parsing the real _ga and _ga_<container> cookie values, rather than minting a fresh client_id, follows from what this guide says those fields do.

  • Validating Measurement Protocol events. https://developers.google.com/analytics/devguides/collection/protocol/ga4/validating-events Supports Chapter 4 and Chapter 6. Documents the /debug/mp/collect validation endpoint and its validationMessages array. The empty validationMessages response is the verification the reconstruction used to prove the corrected payload before trusting it against the live endpoint.

  • User-provided data for the Measurement Protocol. https://developers.google.com/analytics/devguides/collection/protocol/ga4/uid-data Supports Chapter 9 and Chapter 11. Documents the hashed user-provided data fields that underpin identity matching, the same class of hashed email and phone signal the Enhanced Conversions for Leads path depends on.

The consent defaults are a posture for a US-only operator that stays correct if the business ever advertises into the EEA. These references define what Consent Mode requires and how the signal is structured.

  • About consent mode. https://support.google.com/google-ads/answer/13695607 Supports Chapter 10 and Appendix A.8. Establishes that Consent Mode v2 is a Google Ads requirement for advertising into the EEA, the UK, and Switzerland, which is why the property carries region-scoped denied defaults for 32 region codes despite serving only California.

  • Set up consent mode. https://support.google.com/google-ads/answer/13802165 Supports Chapter 10 and Appendix A.8. Documents the basic and advanced consent-mode implementations and the default and update command structure. The paper's two gtag('consent','default',...) calls, the global grant followed by the region-scoped denial with wait_for_update at 500 ms, are an instance of the pattern this article specifies.

G.3 Offline Conversions and the Data Manager API

The offline-revenue loop targets the Data Manager API because the legacy path was foreclosed for new accounts mid-2026. These references document both paths and the cutoff.

  • Changes to offline click conversion uploads (Google Ads Developer Blog, May 2026). https://ads-developers.googleblog.com/2026/05/changes-to-offline-click-conversion.html Supports Chapter 11. Announces the migration of offline click-conversion uploads to the Data Manager API and the block on the legacy path for developer tokens not previously allowlisted, returning CUSTOMER_NOT_ALLOWLISTED_FOR_THIS_FEATURE. This is the hard date, 2026-06-15, that makes the legacy OfflineConversionUpload route unbuildable for a new account and forces the Data Manager API choice.

  • Upload offline conversions (Google Ads API). https://developers.google.com/google-ads/api/docs/conversions/upload-offline Supports Chapter 11. Documents the click identifiers an offline conversion carries, gclid, wbraid, and gbraid, and the conversion-time and value fields. The contact_submissions schema in Appendix B holds one column for each, so that an upload is a projection of a single database row.

  • Import conversions from clicks (Google Ads Help). https://support.google.com/google-ads/answer/2998031 Supports Chapter 11. Covers the business-facing offline conversion import, including the scheduled Google Sheets upload the chapter recommends as the MVP mechanism before any automated pipeline is built, and the 90-day gclid match window the office must upload inside.

G.4 BigQuery Export

The warehouse lifts the retention ceiling at zero cost for this property's volume. These references support the cost claim, the setup permission model, and the schema.

  • BigQuery pricing. https://cloud.google.com/bigquery/pricing Supports Chapter 12 and Appendix A.7. Establishes the free tiers the cost claim rests on: 10 GB of active storage and 1 TB of query processing per month at no charge, with daily batch export free and only streaming export billed. At roughly 5 MB of monthly event data the property sits far inside the free tier.

  • Set up BigQuery Export (Google Analytics Help). https://support.google.com/analytics/answer/9823238 Supports Chapter 15 and Appendix A.7. States the permission model for creating the link: Editor or above on the Analytics property plus Owner on the destination BigQuery project. This is the source behind the corrected access claim in Chapter 15, that the GA4 role was never the binding constraint on the link and that the wall lived in project-side Cloud IAM.

  • BigQuery Export schema (Google Analytics Help). https://support.google.com/analytics/answer/9358801 Supports Chapter 12. Documents the events_ table layout, the nested event_params and user_properties records, and the daily table partitioning. The chapter's reporting queries read this schema.

  • BigQuery Export (Google Analytics for developers). https://developers.google.com/analytics/bigquery/overview Supports Chapter 12. The developer-facing overview of the export, useful alongside the help-center setup article for engineers writing against the exported tables.

G.5 Attribution and Data Retention

Two settings decide whether the property can answer seasonal and channel questions: the attribution model and the retention window.

  • Data retention (Google Analytics Help). https://support.google.com/analytics/answer/7667196 Supports Chapter 12 and Appendix A.4. Documents the two-to-fourteen-month event-data retention range and the nuance that the window governs Explorations and any report using event- or user-scoped custom dimensions, while standard aggregate reports retain effectively indefinitely. This is why raising the property from the two-month floor to fourteen months was not cosmetic for a business whose analysis is cut by service and county.

  • About attribution and attribution modeling (Google Analytics Help). https://support.google.com/analytics/answer/10596866 Supports Chapter 3 and Chapter 9. Covers the attribution models available in GA4 and the conditions under which data-driven attribution operates, including its data-volume requirements and its silent fallback to a rule-based model when a property does not meet them. The maturity model's treatment of attribution as a capability a small property may not yet satisfy follows from this.

G.6 Admin API and Data API

The entire audit and reconstruction ran through two APIs rather than the console. These references define every configuration read and write and every reporting query the paper describes.

  • Google Analytics Admin API v1 (guide). https://developers.google.com/analytics/devguides/config/admin/v1 Supports Chapter 4 and Appendix D. The configuration surface: custom dimensions, key events, data-retention settings, Enhanced Measurement, access bindings, and product links. Every setting recorded in Appendix A was read or written through this API.

  • Admin API v1 REST reference. https://developers.google.com/analytics/devguides/config/admin/v1/rest Supports Appendix D. The method-level reference behind the specific calls the paper names, including customDimensions.create, keyEvents, dataRetentionSettings.patch, accessBindings.list, and bigQueryLinks.create.

  • Google Analytics Data API v1 (guide). https://developers.google.com/analytics/devguides/reporting/data/v1 Supports Chapter 4 and Appendix D. The reporting surface used to read event counts, dimensions, and metrics during the audit, including the runReport method that produced the audit's quantitative findings.

  • Data API v1 REST reference. https://developers.google.com/analytics/devguides/reporting/data/v1/rest Supports Appendix D. The method-level reference for the reporting queries, including properties.runReport and its dimension and metric request bodies.

One technical detail in the paper has no single authoritative Google URL: the internal structure of the _ga_<container> session cookie value, from which Chapter 9 and Appendix B extract the session_id. Google documents the existence and purpose of the _ga and _ga_<container> cookies in its cookie-usage help material, but it does not publish a field-by-field schema of the packed value string, which is an implementation detail of gtag.js rather than a supported interface. The parsing in Appendix B is therefore documented empirically, against the live cookie the property's own tag sets, and flagged in the code as dependent on an undocumented format that a future gtag.js release could change without notice. Where a claim in this paper rests on an observed artifact rather than a published specification, the paper says so; this is the one place in the reference set where that distinction matters.

All URLs verified July 2026. Google first-party documentation only. Product behavior current as of the field work; re-verify load-bearing mechanisms before relying on them in a new build.