Raising Conversion by Measurement Instead of Guesswork

What the Conversion Rate Actually Measures

The conversion rate is the share of visitors who perform a previously defined action, divided by the total number of visitors. It sounds simple, but the argument starts at the definition of the action. A submitted contact form, a purchase, a newsletter sign-up and a tap on a phone number are four entirely different conversions with entirely different value.

Anyone who wants to raise conversion first decides what counts at all. It often pays to separate a macro conversion (the actual business goal, such as a completed purchase) from a micro conversion (an intermediate step, such as adding an item to the cart). Without this clarity the work optimises a metric that nobody truly needs.

A conversion rate without a defined goal is not a metric, it is decoration.

The reference base matters just as much. A rate based on all sessions, on unique users, or on the sessions that actually reached a landing page returns different values. As long as the base stays constant over time, the comparison is valid. If it shifts unnoticed, the comparison turns into apples and pears.

The Funnel as a Map

Conversion rarely happens in a single step. Between first contact and completion lies a chain of stages, and at each one a portion of users drops away. This chain is called the funnel. Making it visible is the fastest way to see where the potential sits.

A simplified example for an online shop with 10,000 visitors a month shows how quickly losses add up.

The overall conversion here is three percent. What matters is not this final figure but the weakest transition stage. Only 50 percent move from the cart into checkout, and that is exactly where the first intervention pays off. Lifting that transition from 50 to 60 percent would yield 360 instead of 300 purchases, a gain of 20 percent at the same traffic. The same improvement at the very top of the funnel would carry far less leverage.

For us this way of thinking belongs to "planning further". First identify the stage with the largest loss and the most realistic gain, then intervene precisely, instead of tweaking everywhere at once. More on this approach is on our Mission page.

Setting Up A/B Tests Cleanly

An A/B test compares two versions of a page under real conditions by splitting the traffic at random. It only becomes clean through discipline before the start.

• Hypothesis first. Not "let us try a green button", but "a higher-contrast button raises the click rate because it is found faster in the layout". A hypothesis can be disproven, a gut feeling cannot.
• One variable per test. If the headline, the button and the image change at the same time, there is no telling afterwards which change worked. Anyone who wants to test several factors at once needs a multivariate test with a correspondingly larger sample.
• Calculate the sample size beforehand. The required number of visitors follows from the current rate, the smallest effect still worth having and the desired confidence level. Without this calculation a test runs either too short or needlessly long.
• Run full weeks. User behaviour on Tuesday differs from Sunday. A test over uneven periods mixes weekday and weekend effects and distorts the result.

What Destroys a Test

The most common mistake is stopping early. As soon as one variant is briefly ahead, the test is halted and declared a success. This is called peeking, and it pushes the rate of false conclusions far beyond the planned five percent. A runtime is fixed in advance and held to, even when the curves wobble in between.

Just as treacherous is running several tests in parallel on the same pages. If the variants overlap, they influence each other, and no analysis stays clean.

The Three Levers With the Greatest Effect

Across industries, three levers work more reliably than any detail tuning of colours or corner radii.

Clarity. Within a few seconds it has to be clear what the offer is and what the next step is. A vague headline costs more conversions than any button that is too small. Concrete beats clever, an understandable statement beats a pun.

Trust. An action that costs money or data demands reassurance. Real references, transparent prices, reachable contacts and a sound data protection basis under GDPR Article 6 lower perceived uncertainty. Trust is not a soft factor, it shows directly in the completion rate.

Reduce friction. Every required field, every unnecessary step and every second of load time costs users. The Core Web Vitals set hard thresholds for this, a Largest Contentful Paint under 2.5 seconds, an Interaction to Next Paint under 200 milliseconds and a Cumulative Layout Shift under 0.1. Cutting a form from nine fields to four often moves the completion rate more than any visual change.

Do Not Over-Read Significance

Statistical significance answers a single question. How likely would the measured difference be if there were in truth no difference at all. At the common 95 percent level a result counts as significant when this probability falls below five percent. That is useful, yet it is routinely stretched too far.

Three confusions appear especially often.

An example makes the first confusion tangible. A variant lifts the rate from 3.00 to 3.05 percent. At a million visitors this difference is statistically significant, in practice it is almost worthless. Two values together are therefore what matter, the statistical certainty and the economic size of the effect, ideally with a confidence interval that shows the plausible range.

What stays robust is to accept a winner only once the predefined sample is reached, the effect is large enough to be worthwhile, and the result is confirmed in a short follow-up observation. Conversion optimisation is not a one-off hit but a planned sequence of tests, in which every confirmed learning sharpens the next hypothesis.

Anyone who wants to measure a funnel cleanly and test the right places will find the way in through the contact form.

---

Where does a funnel lose the most users? We measure the stages and show which lever brings the largest gain.