Google Ads Conversion Tracking with GA4 and GTM

Conversion tracking is not a tag you install at the end of a build; it is a definition you agree at the start. Before anyone opens Google Tag Manager, decide what counts: a purchase, a booked appointment, a form submission that meets a qualification bar, or a phone call over a minimum duration. Then decide which system owns that definition, because Google Ads, GA4, and your CRM will each happily count something slightly different.

In GA4, everything is an event. Some events arrive automatically, enhanced measurement captures others such as scrolls, outbound clicks, and video engagement, and the rest you define to match the business questions the property is meant to answer. Mark the important ones as key events. That marking is the signal that a given event matters, and it is what can be imported into Google Ads as a conversion action.

There is a trade-off between importing GA4 key events and building conversions natively in Google Ads. Importing keeps one definition in one place and avoids maintaining two tagging paths, but it ties the bidding signal to the GA4 implementation and its consent behavior. Native tracking can capture platform-specific signals and often receives enhanced conversion data more directly, but it introduces a second definition that can drift. Pick one owner per outcome and let the other system reference it rather than duplicate it.

The GTM side works best as a contract between the website and the tag setup. Push structured events onto the dataLayer with stable names and the parameters you actually need, such as value, currency, form name, and page context, and document them. Tags then listen for those events rather than scraping the page or depending on CSS classes that change with the next redesign. Preview and debug every new trigger with real interaction, not with a manually fired event.

Consent settings are part of conversion tracking now. Depending on where your visitors are, tags may need to wait for consent before storing or reading identifiers, and consent mode tells Google how to handle the gap. Conversion and behavioral modeling can recover part of the missing signal when consent is denied, but only when the implementation is correct. Treat the consent banner and the tag behavior as one system, and test both states before launch.

Attribution settings determine which clicks get credit. Google Ads and GA4 use their own default models and lookback windows, and they count conversions differently: one may record every conversion while the other records only the first, or attribute a sale to a different channel entirely. Neither is wrong, but reporting from the two systems will never match exactly. Document which number the team treats as the source of truth for decisions, and use the other for diagnosis.

Validation belongs before launch, not after the first invoice. Walk a real test submission through the entire path and confirm the event appears once in GA4 DebugView, once in Tag Manager preview, and once in the Google Ads conversion report. Check that values and currencies pass through correctly, that internal traffic is excluded, and that each conversion action uses the right counting method. Repeat the test after any change to forms, thank-you pages, or the consent banner. If the site has several forms or landing pages, validate a sample of each rather than assuming one working path covers them all.

Double counting is the most common failure. A form that fires on submit and again on the thank-you page, a GA4 import running alongside a native gtag conversion, a purchase event duplicated by both the browser and a server-side tag: all of these inflate results and mislead bidding. Assign each outcome exactly one primary conversion action, mark duplicates as secondary, and check deduplication wherever browser and server events coexist.

Leads that close offline still belong in the measurement loop. If the CRM records which form fills became customers, import those outcomes back into Google Ads using the identifiers available to you, subject to consent and platform policy. Without that feedback, Smart Bidding optimizes toward the leads that are easiest to generate rather than the ones most likely to buy. The same logic applies to calls, where duration thresholds and call quality notes can separate real enquiries from wrong numbers.

Tracking decays quietly. A theme update changes a form, a cookie banner blocks a tag, a new checkout step breaks the dataLayer, and nobody notices until spend looks strange or conversions vanish. Put a recurring check on the calendar: one test conversion per month, a look at the conversion action status column, and a comparison between platform-reported conversions and CRM records. Small discrepancies are normal; a silent zero is not. Assign an owner to that check so the responsibility does not dissolve between marketing, development, and the agency.

Done properly, the setup gives you three things: a definition everyone agrees on, a reliable signal for bidding, and a reporting path you can defend in a budget meeting. The tooling is secondary to those agreements. Google Tag Manager and GA4 only make them operational.