GA4 conversion tracking that does not break

GA4 tracking rarely breaks loudly. It reports a number, the number looks plausible, and nobody discovers it was wrong until a quarterly review or a bidding decision built on it goes badly. Setting it up so that does not happen is mostly about doing three unglamorous things in the right order.

Design the taxonomy before you tag anything

The most expensive GA4 mistakes are naming mistakes, because renaming an event later orphans all of its history. You end up with two events that mean the same thing, a report that has to union them, and a permanent footnote.

• Agree a naming convention and write it down. Lowercase with underscores, verb plus object, consistent across the whole property.
• Decide which parameters every event carries, and which are event-specific. Parameters you forget to collect cannot be recovered retroactively.
• Register custom dimensions for any parameter you intend to report on. An unregistered parameter is collected but not queryable in standard reports, which surprises people months later.
• Keep the event count small. A property with a hundred events usually has ten that anyone reads and ninety that add noise.

Mark only real outcomes as key events

In GA4 a key event is what used to be called a conversion, and it is what feeds Google Ads if you import it. That makes the choice consequential rather than cosmetic.

Newsletter signups, PDF downloads, and video plays are useful engagement signals and they are usually not business outcomes. Marking them as key events and importing them into Google Ads tells the bidding algorithm to buy the cheapest of them, which it will do very effectively.

Attach a value where a value genuinely exists. If a lead is worth a knowable average amount, send it. If it is not, send no value rather than a made-up one, because a fabricated value produces a ROAS figure that looks authoritative and means nothing.

Validate against real interactions

This is the step that separates a setup that holds from one that quietly degrades. DebugView confirms that a tag can fire. It does not confirm that it fires once, on the right action, on the devices your customers actually use.

1. Complete a real submission or purchase on a real phone, on mobile data rather than office wifi, with no debug mode enabled.
2. Confirm the event appears in the realtime report exactly once, with the parameters you expect.
3. Repeat with an ad blocker enabled, and note what is lost. That loss is not random, and it is the argument for server-side tracking.
4. Reload the confirmation page and confirm the event does not fire a second time. Counting set to Every rather than Once is one of the most common inflation bugs.
5. Cross a domain boundary if you have one, and confirm the session survives. A broken cross-domain handoff dumps conversions into direct traffic and makes every channel look worse than it is.

Why GA4 and Google Ads disagree

They disagree by design, and the first thing to establish is that a gap is not evidence of a bug. The two systems use different attribution models, different conversion windows, and different definitions of which interaction gets credit.

• Google Ads credits the conversion to the date of the ad click. GA4 credits it to the date the conversion happened. On a long sales cycle these are different months.
• Google Ads reports on its own attribution within its own window. GA4 attributes across all channels.
• Google Ads can include view-through conversions that GA4 will never see.

Investigate a gap when it changes suddenly rather than when it merely exists. A stable discrepancy is two systems measuring different things consistently. A discrepancy that moved last Tuesday is a tagging change nobody documented.

When conversion tracking stops working

If conversions have dropped to zero or fallen off a cliff, work through the causes in likelihood order rather than starting with the most technical one.

1. Did the site change? A deploy that altered the confirmation page URL or the form submission behaviour is the most common cause by a wide margin.
2. Did the consent banner change? A consent mode misconfiguration can suppress tags for a large share of traffic overnight, and it looks exactly like a tracking failure.
3. Is the tag still on the page? Container changes, cache purges, and CMS template edits all remove tags without telling anyone.
4. Did someone edit the key event definition? Check the change history before assuming a code problem.
5. Is the data merely delayed? GA4 processing lags, and a panic about yesterday's numbers is sometimes just impatience.

The durable fix for all of this is a reconciliation habit: compare GA4 totals against your CRM or order system on a fixed schedule, so a break is discovered by a routine rather than by a bad decision. That reconciliation is a core part of marketing analytics services, and it is what makes google ads management decisions trustworthy in the first place.