Data-driven marketing: measuring is not steering

Data-driven marketing is a way of working in which you base marketing choices on the measured behaviour of visitors, leads and customers rather than on instinct. You decide up front which decision needs to improve, measure only what belongs to it, and check at a fixed moment whether you need to change something.

Almost every marketing team signs up to that definition. Hardly anyone works without Google Analytics any more, the ad platforms deliver fresh numbers every morning, and the CRM knows exactly who got in touch and when. Measurement is rarely what’s missing.

What is missing is the moment after. The report arrives, gets read, and next month’s campaign looks exactly like last month’s. Data-driven marketing is about precisely that moment: not how much you measure, but what changes because you measured it.

Below you’ll find which data you need, how to start in six steps, what an A/B test does and doesn’t tell you at small volumes, and where the limits of privacy and AI lie.

What is data-driven marketing?

It is marketing in which a number has a say in where the next budget, the next campaign or the next test goes. Not the opinion of whoever has been around longest, and not the channel a competitor happens to be advertising on.

Two terms tend to get mixed up here. Data-driven means the number decides. Data-informed means the number weighs heavily, next to what you know about your market and your customers. For most small and mid-sized businesses the second is the honest goal. With forty leads a month a number is rarely strong enough to decide on its own, and if you let it anyway, you’re steering on chance.

That makes data-driven marketing the marketing version of data-driven working: the same order of decision, definition and source, applied to channels, campaigns and leads.

Why data-driven marketing pays off

The first gain is in the budget conversation. Without numbers, the best-defended campaign wins. With numbers, the campaign that demonstrably brings in customers wins. That’s a different conversation, and usually a shorter one.

The second gain is relevance. If you know which pages a visitor read and which question they asked in the contact form, you can tailor the follow-up to it. An email about the topic someone searched for themselves outperforms the general newsletter.

The third gain is speed. A campaign that produces nothing after two weeks gets stopped after two weeks instead of at the end of the quarter, and the money that frees up goes to what does work.

What data do you need?

Less than you think. Start with the questions already on the table, and only then collect data. In most organisations there are three, from three different people:

• The marketer wants to know which channel deserves more budget next month.
• The managing director wants to know what share of new revenue comes from marketing.
• The sales manager wants to know which incoming leads are worth calling today.

Those three questions need four kinds of data. How people arrive: channel, cost, clicks. What they do: which pages they view and where they drop off. What converts: leads, customers and revenue per channel. And what a customer is worth afterwards: repeat purchases and churn.

Everything beyond that is optional. Followers, reach and open rates are easy to measure and rarely the reason anyone decides differently. Which KPIs to choose per department, and how to set a target next to each one, is covered in a separate article.

Data-driven marketing in six steps

1. Pick one decision that needs to improve. Usually that’s how the budget is split across channels. Write the question down before you build anything.
2. Define what a lead and a customer are. Does a downloaded whitepaper count as a lead, or only a request for a quote? When is someone a customer: at the proposal or at the first invoice? One page is enough.
3. Get the measurement right. Conversions set up properly in GA4, consistent UTM tags on every campaign link, and a field in the CRM that keeps the source of the lead all the way through to the order.
4. Let one system count the customer. Every ad platform claims the conversions it played a part in, so added together you have more customers than you actually won. How many conversions a channel is credited with also depends on the attribution model you choose in Google Analytics. Agree that the CRM is the single source of truth, and that the rest provides context.
5. Set a fixed moment. Weekly for the channels, monthly for revenue. Without an agreement, people only look when things go badly.
6. Test one thing at a time. A new ad text, a shorter form, a different subject line. What a test can deliver at small volumes is in the next section.

That order is deliberately the reverse of how it usually goes. Almost every project starts by connecting tools, because that’s the fastest way to get something to look at. Without steps 1 and 2 you are looking, but you don’t know at what.

What does an A/B test tell you at small volumes?

Less than the testing tool suggests. Every source on data-driven marketing recommends A/B testing, and that advice holds for an online shop with tens of thousands of visitors. For a B2B service provider with a few hundred visitors a month on its quote page, it’s a different story.

A worked example. Your quote page converts 2 percent of visitors, and you want to know whether a shorter version reaches 3 percent. To demonstrate that difference with the usual certainty (95 percent confidence, 80 percent power), you need roughly 3,826 visitors per variant, 7,652 in total. At 1,500 visitors a month, that test takes five months.

The temptation is to peek early. After three weeks variant B is ahead by seven requests to five, and someone has already added the new button colour to the brand guidelines. At those volumes, a difference of two requests is simply noise.

What does work at small volumes: test big differences rather than small ones, such as a different offer instead of a different button colour. Test where the traffic is, which is usually the ads. And treat the rest as a decision under uncertainty, not as a proven result. That is exactly the difference between data-informed and data-driven.

Why measuring is not steering

Every Monday the same overview, every Monday the same tile in red, and nobody feels responsible because the number belongs to everyone.

Sound familiar?

Looking at a number costs nothing, and neither does a report full of data insights. Value only appears when someone does something differently because of it: pausing a channel, shifting budget, rewriting a page. Teams that take that difference seriously don’t measure more than others. They simply agree up front what happens when a number moves in a certain direction.

One rule is enough for that: every KPI on your marketing overview has an owner and an agreed response. If LinkedIn leads fall below target two weeks in a row, the marketer decides within a week whether the campaign is adjusted or stopped. It sounds strict, but it prevents a number from sitting in red for months while everyone looks at it.

That only works if the numbers are right. A lead that appears twice in the CRM or was entered without a source weakens every conclusion, and our article on data quality goes into that in depth.

Which tools do you use for data-driven marketing?

For most small and mid-sized businesses the basics are already in place:

• Google Analytics 4 for behaviour on your website.
• Search Console for organic search results.
• The ad platforms (Google Ads, LinkedIn, Meta) for spend and clicks.
• A CRM such as HubSpot or Pipedrive for leads, their source and their status.
• An email tool such as Mailchimp or ActiveCampaign for sends and clicks.

The problem is rarely a missing tool. It’s that each of those systems only reports on itself and stops at the lead. What that lead later turned into sits in your CRM or accounting system. As long as that stays separate, you compare channels on cost rather than on return.

Bringing those sources together on one screen means building a marketing dashboard. Data Studio from Google, called Looker Studio until April 2026, is enough for that as long as you stay within Google. Once revenue and margin from your accounting system need to be included, Power BI is the logical step.

Don’t buy extra tools while steps 2 and 3 of the plan above aren’t in place. Without a recorded source per lead, a more expensive package only gives you a nicer chart of the same guess.

Privacy and AI: where the limits are

Data-driven marketing is about the behaviour of people, and that comes with rules. In the Netherlands, the government guidance on cookies for businesses distinguishes functional, analytical and tracking cookies. Analytical cookies with little impact on privacy may be placed without consent, while tracking cookies need the visitor’s explicit consent beforehand, and simply scrolling on does not count as a yes.

That affects your numbers. Visitors who decline are invisible to you, or only show up as an estimate. So use your analytics for trends, and your CRM and accounting system for the final count.

AI mainly changes the speed. Spotting patterns in campaign data, flagging anomalies, writing variants of an ad text: that now takes minutes. Marketing is also where it’s applied most. Of the European enterprises using AI, 34.70 percent used it for marketing or sales in 2025, according to Eurostat figures. But AI calculates with the data you have. If the source of a lead isn’t in your CRM, even the smartest model won’t find that connection.

---

Will you set next quarter’s budget based on what customers brought in, or based on what happened last quarter too? Get in touch and we’ll look together at which numbers need to be connected for that.

Frequently asked questions