Measurement and attribution

Without reliable measurement, online marketing is an opinion. With measurement it is a discussion about numbers — one you can win or lose, but at least have. Below is how to set up measurement that holds up, now that cookies and privacy rules remove part of the picture.

Start with the question you want answered

Before you set anything up, decide which decisions you want to be able to make. In practice there are four:

• Which channel delivers customers, not just clicks?
• Which pages convince and which do not?
• Where in the process do people drop out?
• What may a lead cost, and what does it cost now?

Anything you measure that answers none of those is reporting without a function.

Record your baseline before you start

Write down where you stand today: sessions per channel, conversion rate, number of enquiries, cost per enquiry and revenue from online channels. Save that as a frozen report, not as a dashboard that later recalculates with new settings. Without a baseline, every later result is an assumption.

The components of a measurement setup that holds up

1. GA4 with deliberate events. Do not measure twenty events but five that mean something: enquiry submitted, phone number clicked, pricing page visited, video completed, purchase.
2. Consent Mode in line with GDPR. Measurement is only allowed after consent; for visitors who decline, Google uses aggregated modelling. A consent banner that hides the decline option is not only legally risky, it also makes your data worse.
3. Server-side or first-party measurement. Measurement running via your own domain is blocked less often. That makes your numbers more stable, but it is not a way to bypass consent.
4. CRM connection. Without feedback from sales you only know how many forms were submitted, not how much money they produced.
5. Search Console. The only source for what people type before they click; indispensable for your content plan.

Why your numbers will never match exactly

Ad platforms, GA4 and your CRM almost never report the same figures. That comes down to three things: different attribution models (Google Ads credits a conversion to the moment of the click, GA4 to the moment of the conversion), different lookback windows, and the share of visitors who give no consent.

A practical agreement: choose one source as the truth for decisions — usually your CRM for revenue and GA4 for behaviour — and use platform numbers only to optimise campaigns inside that platform. Do not chase a 5% difference; do investigate a structural gap of 30% or more.

Attribution without illusions

Last click underestimates everything that creates demand. First click overestimates it. Data-driven models are better, but need volume and explain nothing that happens off your site.

What works best in practice is a combination of three signals:

• Platform data for optimisation within channels.
• A question in your form ("How did you find us?"). Surprisingly informative, especially for hard-to-measure channels.
• Geo or period tests. Turn a channel off in one region or period and see what happens to the total. That measures incremental effect, which no attribution model can.

Which numbers you review monthly

Level	What you look at	Why
Company	online revenue, cost per customer, payback period	tells you whether the whole works
Channel	cost per lead, qualified leads, ROAS	decides where budget goes
Page	conversion rate, exits, scroll depth on key pages	decides what you improve
Search demand	impressions, positions, branded search volume	tells you whether visibility grows

Look at trends over thirteen weeks instead of individual weeks. Weekly numbers in SME businesses often swing more because of one large deal than because of your marketing.

What not to do

• Reporting on sessions and followers. They move with budget, not with results.
• Adjusting every week. You are measuring noise; give a change at least two to four weeks.
• Setting up measurement after the campaign starts. The first weeks are the most instructive.
• Measuring everything. A dashboard with sixty widgets leads to fewer decisions than one page with five numbers.
• Ignoring lead quality. Twice as many leads at half the quality is not growth.

How to set this up practically

Budget one day for the setup (events, consent, goals, Search Console), one day for the CRM connection and then half an hour per month for reporting. That is small work compared to what it prevents: months of spend without knowing what it produced.

For the choices you make with these numbers, read strategy, channels compared and budget and allocation. For a starting split for your mix, use the marketing mix selector.