A Practical Guide to GA4: Reading Your Website Data and Setting Up Conversion Events
You open the dashboard, see a few thousand visits this month, and then what?
The data problem at most company websites is not an absence of numbers but an abundance of numbers that mean nothing. The questions worth answering are quite specific: where did this month’s inquiries come from? Is the blog bringing in customers? Are things better since the redesign? Whether those questions have answers depends entirely on whether the tracking was designed in the first place.
Work out what you want to answer before you configure anything
The most common sequencing error when adopting an analytics tool is to install the tracking code first and only then think about what you want to see.
The more effective approach runs the other way: list the questions you want to answer in a meeting each month, then decide which actions need to be recorded. The questions usually fall into a few categories:
- Which channels are the inquiries or orders arriving through?
- Which pages actually lead to contact, and which are merely viewed a lot?
- How does behavior differ between visitors from organic search and those from ads or social?
- Is newly published content starting to bring in traffic?
Once the list is written, you will find the number of actions you need to record is small — but every one of them has to be defined precisely. This step is the same kind of work as writing software requirements, and the approach in how to write software requirements applies directly: turn a vague idea into a description someone else can follow without getting it wrong.
How GA4 and Search Console divide the work
These two tools are often assumed to overlap. In fact they cover different stretches of the user journey.
| Search Console | GA4 | |
|---|---|---|
| Where it observes | Before the visitor arrives | After the visitor arrives |
| The question it answers | What did they search for, did they see me | What did they do once they were here |
| Typical use | Finding keywords with impressions but few clicks | Comparing conversions and channel performance |
| What it tells you about content | Which topics are worth writing | Which content actually drives action |
The most valuable use in practice is to read the two together. A topic with impressions but poor click-through in search usually means the title and description are not answering the search intent. Traffic that clicks through but takes no further action points either to the content itself or to a page with no obvious next step. The method for choosing topics and structuring them is covered in how to run a brand blog.
One caveat: the interfaces and report names in both tools are adjusted frequently, so this article deals only with the reasoning. For where things sit today, follow the current official documentation.
Events and conversions: defining what success means
An event is a single action a visitor performs on your website. A conversion — called a key event in some interfaces — is one of the few actions you have decided carries commercial value. Skip this step and the reports will show traffic but no outcomes.
For a typical company website, the actions worth recording are roughly these:
- A contact form submitted successfully (note: submitted successfully, not the button being pressed)
- The phone number being tapped (particularly important for mobile visitors)
- A tap on the button that adds your messaging app official account — in Taiwan this is usually a LINE official account, which many businesses use as their main channel for talking to customers
- A catalog, price list, or other file being downloaded
- For e-commerce, add-to-cart, checkout started, and order completed
A few design principles determine how usable the setup will be later:
One: agree naming conventions before you start recording. Once event names are inconsistent, nobody six months later can tell what separates contact from contact_form. Use a fixed pattern such as action plus location, and write it into a document that can be handed over.
Two: mark only genuinely valuable actions as conversions. Marking every event as a conversion is the same as having none, because the reports fill up with numbers that carry no meaning.
Three: distinguish “pressed” from “completed”. A submit button pressed but failing validation, or failing to write on the back end, is not a success. The recording point belongs after the success is confirmed.
Four: tag your traffic sources. Links used in external campaigns, newsletters, and word-of-mouth articles should carry consistent source tags, or they will all be lumped together. The method and examples are set out more fully in how to measure word-of-mouth ROI.
Principles for privacy and consent
Two things have to be dealt with together when you set up tracking.
First, do not send personal data into an analytics tool. Names, phone numbers, email addresses, postal addresses, and national identification numbers should never appear in event parameters or URL parameters. The most common accident is a form that carries the submitted values into the URL afterwards, which records an entire batch of personal data.
Second, notice and consent. Your website should have a privacy policy stating what is collected, why, and how it is retained. If you adopt an approach where tracking is only enabled after consent, make sure the relevant scripts genuinely do not load beforehand, rather than simply displaying a banner that cannot be dismissed.
Regulations and the consent mechanisms offered by each tool keep changing, so what is stated here is only the principle: collect nothing you can avoid collecting, and be explicit about whatever you do collect. For actual implementation, follow the latest versions published by the relevant authorities and the official documentation, and take legal advice where contractual or compliance responsibility is involved. The broader fundamentals of protecting data on a company website are covered in website security basics for businesses.
The most common mistakes
Duplicate installation. A change of vendor, a change of theme, and later a plugin that installs it again — the same page ends up loading two copies of the tracking code, inflating and distorting the data. Check by confirming in the browser that the same identifier appears only once.
The tracking code installed, with no conversions defined at all. This is the most widespread situation, and the result is that three years later you know how traffic rose and fell but not how much business it produced.
Internal traffic not excluded. If visits from your own staff, your development vendor, and test environments are mixed in, the data on a small website is badly diluted.
Settings not carried over during a redesign. The site is rebuilt, every event stops working, and the historical data can no longer be compared. The checks to run before and after a redesign are set out in the website redesign SEO checklist.
Treating traffic itself as the goal. Traffic grows but inquiries do not, which usually means the wrong people are arriving, or the page offers no clear next step.
Ignoring the effect of speed on the data. When a page is slow, visitors leave before tracking takes effect, and the data looks like nobody was interested when in fact nobody got as far as looking. The relevant metrics are covered in the website speed optimization guide.
Different types of website need different things recorded
Brand and service websites. The path to conversion is short, so the emphasis belongs on contact behavior: form submissions, phone taps, messaging account adds. Visitor numbers on this kind of site are usually modest, which makes tagging the source of every single contact all the more important — with a small sample, untagged data shows no discernible difference at all.
Content sites and brand blogs. An individual article does not necessarily produce an immediate inquiry; it builds trust first. Beyond conversions, the things worth watching are reading depth, clicks on internal links, and whether the proportion of readers moving from an article to a service page is changing.
E-commerce sites. The funnel is longer, so record the full path — add to cart, checkout started, order completed — which is the only way to see where people are dropping out. When integrating a payment provider, confirm that the step where the customer returns to your site is recorded correctly, or completed orders will be credited to the wrong source.
Systems with an admin area or member accounts. Behavior after login often involves personal data, so take particular care not to carry account identifiers or names into tracking parameters. Operational metrics for this kind of system should generally come from your own database, with the analytics tool responsible only for the stretch from outside the site to inside it.
Whose job tracking is, and how to write it into acceptance
This tends to be remembered only after a project ends, at which point it belongs to nobody. The better approach is to settle it during the requirements phase:
- Who installs and configures it: the development vendor, the marketing team, or an outside advisor.
- Which events are to be recorded: listed out with the naming convention attached, treated as one of the deliverables.
- How it will be accepted: by walking through the flow in a test environment and confirming that events are recorded and conversions are marked — not merely that the tracking code exists.
- Whose name the accounts are in: owner-level access to the analytics and search tools should sit with your company, not in a vendor’s personal account. This matters as much as ownership of the domain and hosting.
- Handover documentation: a document explaining which events exist and what each one means, so the next person does not have to start from guesswork.
Writing these into the requirements and the acceptance criteria costs almost nothing and avoids having to redo the whole setup later.
What to review regularly once you are live
You do not need to watch the reports daily. Four things once a month is enough:
- Conversions by channel, rather than visits alone.
- Which landing pages genuinely lead to action, which tells you whether content needs reinforcing or reworking.
- Impression and click trends in search, which tells you which topics to write next.
- The difference between mobile and desktop, since visitors in most industries are predominantly on phones and the two often perform very differently.
If your business depends heavily on customers nearby, look at map and local search performance as well — in Taiwan, Google Maps is where a great deal of local discovery starts. The approach is set out in local business marketing and local SEO.
Analytics tools do not generate business. What they do is stop you guessing. Define what success means, tag your sources properly, keep personal data out of the tools, and the judgments that follow finally have something to rest on.
If your website has no tracking yet, or has tracking that nobody can read, talk to NETVANA about your website — we clarify the questions you need answered before discussing how to implement. NETVANA lists performance tracking setup as one of the deliverables in its website development service; for what each service covers, see the software services overview.
Further reading: to make your content genuinely findable in search, see how to run a brand blog; to improve page speed and experience metrics, see the website speed optimization guide; and to turn marketing spend into something you can actually track, see how to measure word-of-mouth ROI. For reporting on the business numbers GA4 does not hold, see Reporting Systems and Dashboards; and for disclosing what your tracking code collects, see How to Write a Website Privacy Policy.
