GA4: Set It Up Yourself or Get It Done Properly?

Installing Google Analytics 4 is not hard. Create a property, get the tag, put it on the site, see numbers appear. Most people can do that in an afternoon and many have.

The problem is that a GA4 property which is collecting data and a GA4 property which can answer a question are two different things, and the first one looks exactly like the second from the outside. Numbers appear either way. That is what makes bad analytics worse than none — you make decisions from it with confidence.

The rule:

Do the installation yourself. Get help with the conversion definitions, and with anything where the answer will change what you spend money on.

What you can comfortably do yourself

Create the property and install the tag. If you are on WordPress, a reputable plugin or a manual insert in the theme header will do it. If you are on Shopify, Webflow or Squarespace, there is a field for the measurement ID. Verify it works in the Realtime report by loading your own site and watching yourself appear.

Set the basics correctly. Data retention defaults to a short window and can be extended — do it on day one, because you cannot retrospectively recover what was discarded. Set the reporting time zone and currency. Turn on the Google Search Console link if you have Search Console. Exclude your own traffic by internal IP so your team’s visits do not pollute the numbers.

Set up enhanced measurement and understand what it gives you. It automatically tracks scrolls, outbound clicks, site search and file downloads. Useful, free, and about ten seconds of work.

Read the four reports that matter. Traffic acquisition, which tells you where people came from. Landing pages, which tells you what they arrived on. Conversions, if they are defined properly. And the comparison between new and returning visitors. Everything else in the interface is available for when you have a specific question.

That is a genuinely useful setup and it costs you an afternoon. Do not let anyone tell you it requires a specialist.

Where DIY setups usually go wrong

Five failures account for most of the broken properties we look at, and all five produce data that appears healthy.

Conversions that are page views of a thank-you page, and nothing else. This is the most common setup and it fails in two directions. It misses anyone who submits a form that redirects elsewhere, and it counts anyone who bookmarks or refreshes the thank-you page. It also cannot see phone calls at all, which for a service business is often most of the enquiries.

No conversions defined whatsoever. The property records traffic and nothing about outcomes. Everyone then argues about whether traffic going up is good.

Cross-domain tracking not configured. If your booking system, shop or form sits on a different domain, a single visitor is recorded as two, and the second one is attributed to a referral from your own site. Whatever channel actually produced the enquiry gets no credit for it.

Referral exclusions missing for payment providers. A customer goes to a payment page and returns. Without an exclusion, the sale is attributed to the payment provider rather than to the campaign that earned it. Paid campaigns look worse than they are, and get cut.

Bot and internal traffic left in. Your own team, your developer, your uptime monitor. On a low-traffic site this can distort everything, and the smaller the site the worse the distortion.

The common thread: none of these throw an error. The reports look plausible. You only find out when a number you trusted turns out to have been measuring something else.

The test that tells you which situation you are in

Try to answer this question from your own analytics: how many enquiries did the site produce last month, and which channel produced them?

If you can answer it in under two minutes with a number you would defend, your setup is fine. Stop reading and spend the money elsewhere.

If you cannot — if the answer is a session count, or a conversion figure you do not trust, or a shrug — then the setup is not doing its job, and every decision you make from it is a guess with a chart attached.

What paying someone actually buys

Not the installation. It buys the definition work, which is the part that requires judgement.

Deciding what counts as a conversion, and distinguishing between them. A newsletter signup and a quote request are not the same event and should not be reported as one number. Getting this wrong makes a campaign that produces cheap signups look better than one producing expensive customers.

Event tracking that survives the site changing. Form submissions tracked properly rather than by thank-you page. Phone taps on mobile. Clicks on your email address. Booking completions.

Tag Manager set up sanely, so future changes do not require a developer and do not silently double-count.

Verification. Somebody actually submitting the form and confirming the event fires once, correctly, with the right parameters. This step is skipped constantly and it is the entire point.

Linking to Google Ads with the conversions imported correctly, so that campaign bidding optimises towards enquiries rather than clicks. If you spend on Google Ads, this alone usually justifies the work — the platform’s automated bidding is only as good as the signal you feed it.

A report that says something. Not the default dashboard. A short view showing enquiries by channel and by landing page, in language your team understands, so the data gets looked at rather than ignored.

The honest recommendation

If your site gets modest traffic and you are not spending on advertising, install GA4 yourself, define one conversion for your main contact form, and get on with your business. You do not need us and paying for a full measurement setup would be premature.

The moment that changes is when you start spending money based on what the data says. Advertising budget, content investment, a decision to drop a service because “the page gets no traffic” — at that point the cost of being wrong exceeds the cost of getting it right, and the calculation flips.

The worst position is the middle one: spending on marketing while trusting numbers nobody has ever checked. If that describes you, the fix is usually a few hours rather than a project.

What we do here is set out on the web analytics page. If you would rather just find out whether your current setup is telling you the truth, ask us and we will look.