Marketing measurement: four numbers that tell the truth

Marketing measurement usually falls down on the same thing: people track numbers that look good in a report but say nothing about the business. Impressions, likes and visitor counts are typical examples.

In this guide I go through which metrics are genuinely worth tracking, how to set them up and how to calculate marketing’s return so the figure survives scrutiny with your accountant too.

Vanity metrics and what to replace them with

Vanity metric	What to track instead
Impressions	Clicks and their cost
Visitor count	Conversions and their source
Likes and followers	Enquiries from the channel
Click-through rate alone	Cost per enquiry
Ranking for one keyword	Development of organic conversions

The rule is simple: if an improvement in the metric does not mean more sales, it is report filler. It may be a useful intermediate signal, but it is not a goal.

Four metrics that are enough for most

1. Cost per lead (CPL)

A channel’s cost divided by the enquiries it produced. This is the first number that tells you how channels compare. If Google Ads produces an enquiry for €40 and Meta for €90, you know where the next euro belongs, unless the quality differs.

2. Customer acquisition cost (CAC)

Marketing cost divided by customers gained. This matters more than CPL, because cheap leads can be bad leads. A channel producing expensive but buying enquiries is better than one producing cheap but useless ones.

3. Customer lifetime value (CLV)

How much margin the average customer produces over the whole relationship. This figure decides how much acquisition may cost. If a customer produces €3,000 of margin and acquisition costs €300, the right move is to acquire more, not to economise.

4. Return on marketing (ROI or ROAS)

The sales or margin a campaign produced relative to its cost. In ecommerce ROAS (sales divided by ad spend) is straightforward. In services, calculate on margin rather than revenue, or the figure misleads.

How to set measurement up

1. GA4 and conversion events. Define events for form submissions, phone clicks, email clicks and purchases. Without these, analytics reports traffic, not results.

2. Campaign tags on links. Use UTM parameters in all ad, email and social links. Without them traffic ends up in miscellaneous buckets and you cannot tell which channel produced what.

3. Ask leads where they heard of you. One field on the contact form or a question at the start of the call. This is the only way to measure word of mouth, offline advertising and, increasingly, AI search: the answer “I asked ChatGPT” is becoming common fast.

4. Connect the sales data. Record in the CRM or your notes which channel each sale came from. Without this you know the price of leads but not which channel produces real customers.

Attribution, or who gets the credit

The customer journey is rarely straight. The same person sees an Instagram ad, later searches for the company by name in Google, reads a blog article and gets in touch directly a week later. Which channel produced the sale?

The last-click model gives the credit to the branded search and undervalues whatever created the awareness. The practical answer for a small business: do not try to build a perfect attribution model. Track channel-level figures as indicative, ask leads for their source and look at the whole: when advertising is paused, does the total number of enquiries fall? That is often the most honest metric.

Reporting people will actually read

A good monthly report fits on one page and answers four questions: what was done, what it cost, what it produced and what happens next. Everything else is an appendix.

Avoid a report consisting of dozens of charts with no interpretation. A number without a conclusion does not help decision-making. Every metric should be followed by a sentence about what will be done about it.

How often to measure

Staring at it daily leads to overreaction: a single day’s variation is randomness, not a trend. The rhythm that works is a quick weekly glance for anomalies, a proper monthly analysis and decisions, and a quarterly wider review of whether the channel mix is still right.

Remember the lag too: search engine optimisation and content take effect with a delay of months, so they cannot be judged on the same timescale as search advertising.

The GA4 basic reports worth knowing

GA4 is broad, but a small business needs four views from it in practice.

Acquisition. Where visitors come from by channel. This is the first place you see what works and what does not.

Pages and screens. Which pages interest people. Connect this to conversions: a well-read page producing nothing needs a call to action.

Events and conversions. How many forms, calls or purchases arrived and from which channel.

Comparisons. The current period against the previous one or the same period a year ago. A single figure says nothing without a point of comparison.

Alongside these it is worth building one custom summary report with exactly your own metrics. It saves time every month.

How traffic from AI search shows up

Traffic from AI services is recorded in GA4 as a referral (chatgpt.com or perplexity.ai, for instance), but in the default report it disappears among the other referrals.

It is worth creating a custom channel group or segment that gathers those sources together. That way you see how much AI visibility brings visitors and, more importantly, how well those visitors convert. Experience suggests they often convert better than average, because they already arrived with a recommendation.

Some AI traffic does not show at all: the user reads the answer, searches for the company by name and arrives directly. That is why the question “where did you hear about us” is still necessary.

Seasonality and comparing to the right period

A common mistake is comparing a month against the previous month in an industry with strong seasonality. July looks like a catastrophe and November like a miracle, though both are normal.

So compare against the same period the previous year, and look at the trend over a longer stretch. A three-month rolling average shows the direction more reliably than an individual month.

Summary: building the framework from scratch

If there is no measurement at all yet, build it in this order. Get GA4 and conversion events in place first, because nothing else works without them. Then add UTM tags to all ad and email links.

Third, add the question “where did you hear about us” to the contact form. It fills in analytics’ blind spots, of which there are more than most people think.

Finally calculate the value of a customer and how much one new customer may cost. That figure is the foundation of the whole framework: without it you cannot say whether a €60 lead is cheap or expensive.

Frequently asked questions

What is worth measuring in marketing?

Four numbers are enough for most: cost per enquiry, customer acquisition cost, customer lifetime value and return on marketing. Impressions, followers and visitor counts are intermediate signals, not goals.

How do I calculate marketing ROI?

Subtract the marketing cost from the margin marketing produced and divide by the cost. Use margin rather than revenue, or the figure looks better than reality.

What is a good cost per lead?

One that is clearly lower than the margin a customer produces multiplied by the share of leads that become customers. Industry benchmarks are misleading, because customer value varies enormously.

How do I measure what does not show in analytics?

Ask every new customer where they heard about the company. That is the only way to capture referrals, offline visibility and customers arriving from AI search, which does not record reliably in analytics.

How often should marketing be reported?

Monthly is enough for decision-making. A weekly glance for anomalies is useful, but daily tracking usually leads to overreacting to random variation.