Checking Google Ads conversion tracking: a checklist for your audit

A conversion tracking audit does not only check whether a tag fires. It checks whether Google Ads is optimising towards the intended business outcome: which state counts, whether it occurs exactly once, which campaigns bid on it, and whether the result can be checked against the source system. Which audit steps are possible depends on the access and evidence available; access to Google Ads alone does not show the entire measurement chain.

Which business outcome counts as a conversion?

Start with the business transaction, not with the tag manager. In a shop, that is the confirmed order with a unique order ID. In B2B lead generation, it is the valid enquiry stored in the CRM or in the application database with its own ID. A button click, a form start or a thank-you page view sits before or after that state and can be repeated; it is a diagnostic signal, not the main conversion. A browser confirmation alone is generally not sufficient; the exception described in the FAQ – a verified one-to-one representation of the stored business event – remains possible.

The inventory list records, for each field, what is defined and what is still to be checked. An empty field identifies a check still to complete. An implementation is recorded as absent only when its absence has been verified; otherwise the state stays unknown or not auditable:

“Not auditable” and “open” must be read neither as “working” nor as an observed zero.

Which conversion actions influence bidding?

Google Ads groups conversion actions into goals. Primary actions appear in the “Conversions” column and are used for bidding provided the standard goal they belong to is used for bidding. Secondary actions serve observation and appear in “All conversions”.

The exception matters for the audit: a secondary action that is included in a custom goal is nevertheless used for bidding provided the campaign actually uses that custom goal for bidding. “Secondary” on its own therefore does not prove that an action has no influence on the bidding logic. For each action, therefore, check the name, source, business meaning, role, goal membership, the campaigns using it, counting method, value, window and status – and then every custom goal individually.

Micro-actions such as scroll depth, page view or form start are useful for UX diagnosis. If they are used for bidding, the platform optimises towards easily reached intermediate steps instead of the outcome.

What do “One” and “Every” do – and what do they not do?

“Every” counts all conversions after an ad interaction and suits repeat purchases. “One” counts one conversion per action per ad interaction; Google describes this as the choice for leads where only one unique lead per click creates value.

The setting does not replace application-level control. If the same enquiry is stored twice because of a network retry, two business transactions exist, even if Google reports one per click. The result only becomes reliable with a suitable counting method, server-side duplicate protection, a deduplicated import and reconciliation.

Does the tag fire on the right event?

Test the real user path and evaluate at ID level. A green tag status proves nothing about whether the right business event was recorded. Google documents action states such as “Inactive”, “Unverified” and “Needs attention” as well as a report refresh of 3 to 24 hours. A test conversion via Tag Assistant checks that the tag fires, not what it means.

Keep two things apart that are easily conflated: the stored business transaction in your own system and the permission to transmit a conversion from it to Google Ads. The first is a business fact, the second a consent and legal question. Both appear in the log in separate columns:

The table describes isolated test cases; it counts the additional transactions or conversions arising from each case. The 0/1 values in the stored-transaction column hold at the level of the stored event, while the expected count in Google Ads reflects the defined counting method, an eligible attributable ad interaction, permitted successful transmission and reporting latency: successful delivery alone does not prove a visible report count of one, and data-driven attribution can also split credit across several ad interactions. Case 3 is a critical regression case: in the described example the browser shows a success message although no business transaction is created; anyone who counts browser confirmations counts a conversion here that never existed in business terms. Case 10 illustrates the reverse situation – the conversion remains valid even if the email delivery fails. Case 11 illustrates an important distinction: consent granted after the fact does not automatically turn a stored transaction into a conversion that may be sent to Google Ads. For repeat cases 5 and 6, zero means no additional transaction and no additional conversion, not deletion of the original one. Whether and from when a click ID and an import are permitted follows from a documented consent and legal rule, not from the click on the banner.

For each case, record the transaction ID, duplicate key, storage state, server response, row count in the source system, transmission decision, import attempt ID, Google's response and retry status. The test must fail visibly on any deviation.

How maitiq checks the measurement chain on an ongoing basis

In the free audit, maitiq reads the connected account or your Google Ads exports, changes nothing and shows for each check whether it was completed in full, with limitations or skipped; the ongoing managed service is a separate, separately agreed service. Supplied CRM data can be prepared and checked as part of the agreed work – this hashes identifiers, not whole CRM outcomes; a live CRM conversion upload to Google Ads is not currently available through maitiq. The audit itself performs no upload and no Google Ads modification. As a standard, reporting covers the last seven completed days in the account time zone; supplied audit data may restrict the period, and the user-facing reporting portfolio also allows selected periods. Without conversions, or when the denominator is unavailable, the CPA stays empty rather than zero. In manual mode, you decide on each proposal using the proposed action, rationale and supporting data. A recorded decision to exclude exactly the reviewed keywords can also authorise their implementation. When explicitly enabled, the limited autopilot can approve eligible proposals within its configured permissions and limits and apply them through the controlled, logged implementation process.

Are there duplicate sources for the same outcome?

Duplicate conversions arise when the same lead arrives through a direct tag, a GA4 import and an offline import. Segment by action and source. If two actions measure the same outcome, designate exactly one source as used for bidding; the other remains a secondary source for cross-checking and must be excluded from the main conversion – including from all custom goals.

An illustrative starting setup:

• one server-side conversion from the source system as the only main conversion used for bidding;
• the GA4 view exclusively as an additional control analysis;
• never both as primary;
• manual or click-oriented bid control as a possible interim choice until the import is verified, subject to campaign support and the account strategy.

Which import method is supported and permitted?

The import route is itself an audit question. For offline conversion imports and Enhanced Conversions for Leads, Google points to the Data Manager API and describes transitions between the interfaces. Check in the specific account and against the current official documentation which route is supported, which conditions apply and which deadlines are actually in force; do not adopt outdated dates from older guides.

Enhanced Conversions for Leads requires hashed user data; Google recommends supplying the available GCLIDs and requires them when no tag collects that data. Both touch on the legal basis, consent and retention and therefore belong in the audit as a separate check.

Why do Google Ads and GA4 differ in their timing?

Google Ads reports the standard conversion columns by the date of the ad interaction – usually the click, for some conversion types also a view – not by conversion time. A conversion on a Tuesday therefore appears in the previous week if the relevant interaction fell there. There are also columns by conversion time, which Google explicitly names for comparison with third-party analytics.

Never compare daily totals blindly. Record the action, time zone, click or conversion time, attribution model, window, consent, filters and import latency. The definitions of these columns and windows are described in the metric and attribution definitions. For GA4 imports, Google states up to 24 hours until visibility and notes that reports may contain modelled conversions. A difference is a reason to investigate, not automatically an error.

Which attribution model is being used?

Google Ads currently supports two models: data-driven and last click. First click, linear, time decay and position-based are no longer supported; affected actions have been switched to the data-driven model. Check therefore whether such a model change happened unnoticed in your account and whether your reporting logic still matches it.

The model assigns shares of the recorded outcome to the ad interactions involved. It does not repair a wrongly defined conversion. Clarify the action, deduplication, tag and import first. The lookback window of the attribution reports and the conversion window remain separate settings.

What consent settings should the audit check?

Consent determines which signals may be stored and sent. Tag Manager documents ad_storage, ad_user_data, ad_personalization and analytics_storage as well as optional additional consent checks per tag. Check the default state, the update after the user's decision, regional handling, the removal of ad-click identifiers when consent is denied (depending on the configured feature), click ID behaviour, tag firing after refusal, the privacy policy, retention, access and debugging. This check covers technical signal behaviour and permitted transmission; consent mode itself is not a legal determination.

Click IDs and enhanced conversions must not be added as an afterthought; they require explicit legal, consent, retention and security decisions, which appear in the audit as a separate check.

How is the reconciliation against the business source performed?

Compare IDs and states, not just totals. An import without a source ID and status cannot be explained. An email hash is not a deduplication key: it does not answer whether this is the same transaction or a new enquiry. Downstream steps such as appointments remain a metric of their own, because one enquiry can trigger several of them.

What is the audit result?

The result is a list of findings ordered by priority, with the supporting evidence, not a new tracking setup. The recommended template for each finding records the observed state, the expected definition, the evidence, the impact, the affected campaigns, the responsible party, the proposed change, the test, the reconciliation and a recovery plan where feasible. Each item is recorded where it is supported or applicable and otherwise marked explicitly as unknown or not applicable; a proposed test is therefore not presented as executed, and a suggested recovery plan is not presented as verified. The audit delivers its findings within the available evidence and the stated coverage. Implementation follows only after your approval; the audit itself remains read-only. How a finding becomes an approved, verified and – where feasible – reversible intervention is described in the approval process for Google Ads changes.

Frequently asked questions

Sources and how to read them

The sources on this page are documentation from Google Ads, Google Tag Manager and the Google developer documentation on consent mode. They explain platform features and limits. Details of how maitiq works can be found at maitiq.com.

Access to Google Ads alone does not establish the entire measurement chain. For website tracking, CRM quality or revenue attribution, the relevant systems and evidence are included only when they are separately agreed and provided; which checks are possible depends on that access.

Have maitiq review your specific case.