How to Set Up Goals and Conversions in GA4
Everything else in analytics measures activity. This is the only part that measures outcomes. A property recording nothing but visits is answering the wrong question, however carefully anybody reads the reports it produces.
Without This, Nothing Else Matters
Ask a business how many enquiries its website produced last month and most cannot answer. They can tell you visits, pages and which town people appeared to be in. None of that says whether the website did any work. The reason is almost always that nobody ever configured what counts as a result.
What an unconfigured property tells you. Activity only.
How many people arrived and what they looked at. Every figure describes movement rather than outcome. No amount of careful reading turns one into the other.
Why it is the commonest gap. Nothing prompts it.
Installation produces working reports immediately. There is no warning that the important half is missing, so it never reaches anybody's list.
What it costs. Every decision after it.
Whether the new pages work, whether search is producing anything, whether the money spent last quarter did anything. All unanswerable without this.
Why waiting is expensive. No back history.
Configure it today and you have data from today. The comparison you will want in a year begins now rather than retrospectively, which is why we put it in the setup guide too.
What the missing half looks like commercially. An argument you cannot win.
A supplier says the work is producing results. You suspect it is not. Without outcome data neither of you can settle it. The discussion then runs on traffic figures that suit whoever is presenting them.
How long it takes. An afternoon at most.
For a business with a form and a telephone number, considerably less. The thinking in block four takes longer than the configuring.
Everything Is An Event
The current version records everything that happens as an event. A page opening is an event. A video playing is an event. A form being sent is an event. That single model replaced the older arrangement where pages and visits were the central units. It is why the tool feels unfamiliar to anybody who used the previous product.
Why it changed. Websites stopped being pages.
People now do a great deal without loading anything new, so counting page loads had become a poor description of activity. Counting events handles both kinds of site.
What gets collected automatically. A useful amount.
Pages being viewed, visits starting, files being downloaded and outbound links being clicked are among the things recorded without anybody configuring them.
What has to be defined. Anything specific to you.
A particular button, a particular form, a particular step in your booking process. The tool cannot guess which parts of your site matter commercially.
Why this matters before you configure anything. Some of it is already there.
Businesses build elaborate tracking for things already being collected. Checking what exists first saves the work. Frequently the raw material for your first conversion is already sitting in the property.
Where the automatic collection helps most. Small sites.
A five page brochure site with a form and a telephone number may need almost nothing configured beyond marking what is already recorded. The complexity scales with the business rather than being fixed.
What we will not print. Any of the code.
No snippets, no parameter names and no configuration examples anywhere in this cluster. Those date quickly and they belong with whoever maintains your site.
A Conversion Is Just An Important Event
There is no separate conversion system. You take an event the tool is already recording and mark it as important. It becomes a result. That is the entire mechanism. Once it lands, the subject stops being intimidating.
What marking does. Promotes it.
The event begins appearing in reporting as an outcome rather than as one more thing that happened. It becomes available to the reports that answer commercial questions.
What it is called now. A key event.
The interface renamed conversions to key events in March 2024. The idea did not change. Readers arrive searching for one word and see the other, which accounts for a good deal of confusion.
Why the distinction is a business one. You decide.
Nothing technical determines which events deserve marking. It is a judgement about what your business considers a result, which is why block four comes before any configuration.
How many to mark. Few.
Marking everything defeats the purpose, since a report where twenty things are important contains nothing important. Most small businesses need a handful.
What to do with the ones you do not mark. Leave them recording.
An unmarked event is still collected and still available if you change your mind. Marking is about what appears as an outcome rather than about what gets stored, so there is no cost to being selective.
What changes in the reporting. The question you can ask.
Rather than how many people came, you can ask how many did the thing you care about and where those specific people came from.
Decide What Counts Before You Build Anything
Most businesses configure whatever is easy to measure rather than what matters. That is how properties end up counting newsletter sign ups while the telephone rings unrecorded. The list should come from how the business actually gets work.
Start with a pen. Not the tool.
Write down every way a customer currently reaches you, in order of value. That takes ten minutes and it is the specification for everything that follows.
What usually tops the list. The telephone.
For most trades and service businesses, calls outnumber every other route. That is inconvenient for measurement and it does not change where the work comes from.
What comes next. Forms and email.
Contact forms, quote requests, booking enquiries and direct email. These are the routes that record cleanly, which is why they dominate most properties whether or not they dominate the business.
The trap in the list. Measurable is not important.
A brochure download is easy to record and rarely means anything. A telephone call is hard to record and frequently means everything. Configure by importance and accept the gaps.
What to exclude deliberately. Soft actions.
Scrolling, time on a page, visiting the contact page without doing anything. Marking those as results inflates your reporting and makes the genuine outcomes harder to see.
The question that settles most arguments. Would you pay for it.
If somebody offered you a hundred of these tomorrow at a price, would you take them. If the answer is no, it is not a result and it should not be counted as one.
Where this list also gets used. Everywhere else.
The same list drives your advertising measurement, your monthly reporting and any dashboard anybody builds. It is worth an hour rather than ten minutes.
Form Submissions
For a form submission to be recorded reliably, something has to signal that it was actually sent rather than merely opened or abandoned. How straightforward that is depends entirely on what your form does after somebody presses the button.
The arrangement that works cleanly. A thank you page.
A form that sends people to a separate confirmation page after submitting gives you an unambiguous signal, since arriving at that page means the form completed.
The arrangement that causes trouble. A message in place.
Forms that show a confirmation without changing the address are extremely common on modern sites. Nothing distinguishes a successful submission from somebody sitting on the page, so nothing gets recorded.
Why that matters more than it sounds. Silent failure.
The property reports zero enquiries while the business receives them daily. Everybody assumes the website is failing when the measurement is.
What fixing it involves. A developer, usually.
Either changing the form to use a confirmation page or arranging for the submission to signal the tool directly. Both are small jobs for whoever maintains the site.
The check worth doing today. Submit your own form.
Send a real enquiry and see whether anything appears. That two minute test settles whether you have this problem. A surprising number of businesses discover they do.
The version that catches larger businesses. Several forms, one owner.
A site accumulating forms over years, added by different people, rarely has all of them recorded. The newest and the oldest are the usual casualties. Nobody notices, because the total still moves.
The related trap. A thank you page people can reach directly.
If that page is linked from anywhere or gets shared, arrivals will be counted as enquiries. Worth checking that nothing points at it.
Phone Call Clicks
A click on a telephone number can be recorded. An actual call cannot. That distinction is the largest measurement gap in local marketing and it affects exactly the businesses this cluster is written for.
What can be captured. The tap.
Somebody tapping a number on a mobile is an action on your page, so it can be recorded and marked as a result. That is genuinely useful and it is worth configuring.
What that tap is not. A call.
It does not tell you whether the call connected, whether anybody answered, how long it lasted or whether it produced work. Everything after the tap happens outside the website.
Who is completely invisible. The readers.
Anybody on a desktop who read the number and dialled it. Anybody who wrote it down. Anybody who rang the next morning from a different phone. None of them exists in this reporting.
Why that is not a minor caveat. It is most of the enquiries.
For a plumber, a roofer or an electrician the majority of work arrives by telephone. So a property measuring only forms is reporting on the smaller half and calling it the total.
What actually closes the gap. Asking.
Ask every caller how they found you and write it down. It is unglamorous, it costs nothing and it is more reliable than anything a website can infer.
On dedicated call tracking. Handle with care.
Arrangements exist that record calls properly and we name no products. There are consequences for a local business showing a different number in different places. our guide to tracking calls from a business listing covers the caution in full.
Downloads
File downloads are recorded automatically in the current version, so this needs less work than the other two. Whether it belongs among your results depends on whether a download actually indicates anything about somebody's intention to buy.
Who it genuinely suits. Considered purchases.
Businesses whose customers take specifications, price lists or technical documents away to compare. A manufacturer, a supplier or a professional practice with a detailed brochure.
Who it does not. Most local trades.
If your customers ring within a minute of finding you, a download is not part of the process. Marking it as a result adds noise to reporting that needs clarity.
What a download actually signals. Interest, not intent.
Somebody took a document. They may be a customer, a competitor or a student. Treating that as equivalent to an enquiry overstates your performance.
How to use it sensibly. As a secondary measure.
Watch it, do not mark it as a result, then look at whether people who download also enquire. That relationship tells you whether it deserves promoting.
Why it was folded into this page. Demand.
It is the most niche of the tracking subjects and it does not need a guide of its own, which is a decision about your time rather than an oversight.
Not Every Conversion Is Worth The Same
Counting enquiries without weighting them treats a tap washer replacement and a full bathroom installation as identical results. A month with thirty small jobs and a month with eight large ones look completely different to the business and identical in the reporting.
What assigning values does. Restores proportion.
Each kind of result carries a figure, so the reporting reflects what a month was actually worth rather than how many times something was clicked.
How accurate the figures need to be. Roughly right.
An average job value per enquiry type is enough. The purpose is proportion rather than accounting. Nobody is reconciling this against your books.
How to work them out. Two numbers.
What a job of that type is typically worth and roughly what share of those enquiries become work. Multiply and use that. It will be wrong in detail and right in shape.
What it changes. Which sources matter.
A source producing few but valuable enquiries can outrank one producing many small ones. Unweighted counting hides that completely and misdirects spending.
The caution. Never report it as revenue.
These are estimated values attached to enquiries rather than money received. A figure like that appearing in a report next to actual turnover invites a comparison it cannot survive.
When not to bother. One kind of job.
A business doing essentially the same work at a similar price every time gains nothing from values. Counting the enquiries is already proportional. Adding figures is effort for no additional insight.
Where the real thing lives. Ecommerce tracking.
Shops record actual transaction values, which is a different and larger job. Our ecommerce guide covers it, including why that figure will not match the till either.
Test It Before You Trust It
A conversion configured and never verified is worse than having none at all. No tracking leaves you knowing you cannot answer the question. Broken tracking leaves you confidently answering it wrongly. Nothing about the reporting looks unusual.
What to do. Be a customer.
Send a real enquiry through every route somebody might use. Fill in the form, tap the number on a phone, send the email. Then check each one arrived.
Why every route. They fail separately.
A site with three forms frequently has one of them unrecorded, usually the one added later by somebody else. Testing the main one proves nothing about the others.
What failure looks like. A believable number.
Not zero, which somebody would notice. A plausible figure that happens to exclude half your enquiries, which nobody questions because it looks reasonable.
When to retest. After anything changes.
A redesign, a new form, a plugin update or a platform change can all break this silently. Add it to whatever checklist follows a site change.
The sanity check that costs nothing. Compare against reality.
Count the enquiries you actually received last month and compare. If the reporting says four and your inbox says eleven, you have found something.
Where this leads next. Reading it properly.
Once results are recorded, the rest of this series becomes considerably more useful. Everything sits on the Google Analytics guide. Our measuring performance material covers what to do with outcome data.
How many
enquiries did
your website
produce last month?
Most businesses can tell you visits, pages and which town people appeared to be in. None of that says whether the website did any work. Nothing warns you that the important half was never configured, which is why properties run for years producing reports that answer the wrong question.
What we configure and verify:
If the reporting says four enquiries and your inbox says eleven, the website is not the thing that is broken.
Every guide.
One tool.
Getting it installed and configured properly, reading the reports without being misled, measuring outcomes rather than activity, the advanced reporting and what to do when the numbers look wrong.
