Tracking Conversions in GA4 Without Guesswork

Paste the tag, see the numbers, assume you are measuring. That assumption is where most analytics work goes wrong. A brand-new Google Analytics 4 property with the base tag installed will happily report users, sessions and page views from day one. None of those tell you whether anybody contacted you.

This is not a criticism of the tool. It is a description of how it ships. The measurement that matters has to be specified by somebody who knows what a valuable action looks like on your site, and nothing in the default install knows that.

What a fresh property genuinely measures

Out of the box you get automatically collected events plus a bundle called enhanced measurement, which you can inspect under Admin and then your data stream’s settings. Enhanced measurement is where page views, scrolls, outbound clicks, site search, video engagement and file downloads come from. It is a reasonable set of defaults and it is worth leaving on.

It is also, for a service business, close to useless as a measure of commercial outcome. Knowing that someone scrolled ninety per cent of a page tells you almost nothing about whether the phone rang. Enhanced measurement includes form interaction events too, and they deserve caution: they fire on generic form behaviour rather than on your specific enquiry form, so on sites with a search box, a newsletter field and a contact form they cheerfully lump all three together.

So the honest starting position is this. A default install measures attention. It does not measure demand.

Events, key events, and the difference that matters

Everything in GA4 is an event. A page view is an event. A click is an event. Nothing about being an event makes something important.

Importance is applied afterwards by marking an event as a key event — the label that replaced the older word “conversion” in the reporting interface. In Admin you will find the events list and the key events list grouped together under data display; Google has shuffled the exact position of these screens more than once, so if the labels have moved since you last looked, they will still be adjacent in the same admin group.

Two routes exist. If the event is already being collected, find it in the events list and switch on the toggle that marks it as a key event. If it is not being collected yet, you can create the key event by name in advance, and it will start populating once the event begins arriving. The name has to match exactly, character for character, and this is the single most common reason a key event sits at zero for a fortnight before anyone notices.

One important limit: marking an event as a key event is not retroactive. Data collected before you flipped the switch stays unclassified. Whatever you set up today starts counting today.

Decide what counts before you build anything

Skip this step and you end up with fourteen key events, none of which anyone trusts.

Write down the actions that genuinely represent a person choosing to speak to you. For most businesses that list is short:

• An enquiry form submitted successfully — not started, not attempted, submitted.
• A click on a phone number link on mobile.
• A click on an email address link.
• A booking, quote request or checkout completion, if you have one.

Then write down the actions that feel valuable but are not. Newsletter signups, PDF downloads and brochure views are engagement, not enquiries. They can be tracked as ordinary events and looked at occasionally. They should not sit in the same column as the things that produce revenue, because the moment they do, the total becomes meaningless and every report built on it inherits the problem.

A useful discipline: if a number went up and you would not change any decision as a result, it is not a key event.

Three ways to wire it up

The thank-you page. If your form redirects to a distinct confirmation URL, this is the most robust option available. A page view of that URL is a real submission, because the server only serves it after processing. Create a custom event triggered on that page path, mark it as a key event, and you are done. It also survives site redesigns better than click-based tracking.

A tag manager trigger. If your form submits without changing the URL — most modern forms do — you need to listen for the success state. In Google Tag Manager that usually means a trigger on the form’s success message appearing, or on a dataLayer push that your developer adds when the submission succeeds. Both are more reliable than a plain click trigger on the button, which fires whether or not the form validated.

Direct tagging. Your developer fires the event from the site’s own code at the moment the submission is confirmed. This is the most accurate route and the least convenient one, because it means a code change every time the definition shifts. It is the right answer for anything involving a genuine transaction.

Where you have the choice, use Google’s recommended event name for lead generation rather than inventing your own. Standard names carry built-in reporting behaviour and import cleanly into ads platforms; a name you made up does neither.

The checks that catch broken tracking

Real-time reporting is not a verification step, it is a smoke test. Do these instead.

Submit the form yourself, from a phone, on the live site. Watch the event arrive with the parameters you expected. Then do it again with a deliberately invalid field to confirm the event does not fire on a failed submission.

Compare your key event count against the actual inbox for the same month. If GA4 says thirty-one and the inbox holds nineteen, something is double-firing — usually a tag placed both in the template and in tag manager. If GA4 says nine and the inbox holds nineteen, consent handling or an ad blocker is eating a share of them, and you now know the size of the gap.

Set a recurring reminder to repeat that reconciliation quarterly. Tracking does not break loudly. It breaks during a theme update, and nobody notices until a quarter of reporting is already wrong.

What it still will not tell you

GA4 measures what happened in a browser. It does not know that the enquiry from a Tuesday was a time-waster and the one from Thursday became your largest job of the year. It will not see a call placed from a number someone read off the page and dialled by hand. It will not see the customer who found you online and walked in.

Those blind spots are not failures of setup. They are the boundary of the tool. Accurate measurement inside that boundary is still worth having — it is the difference between deciding where to spend and guessing — but the numbers should be treated as a strong signal rather than a complete account.

If your property has been collecting data for months and you are not sure any of it is trustworthy, an audit of what is actually firing is the sensible first step. That is where our web analytics work usually starts, and it is worth doing before you spend anything more on ads.