What is conversion rate: calculation and improvement

The short answer to what is conversion rate is this: the share of visitors in a given period who completed a defined action, out of the total visitors in that same period. Because the formula is this simple, most teams skip past calculating the number and never discuss what the number measures. Yet it is possible to calculate three different conversion rates for the same site in the same week, and all three are technically correct. The difference comes from the denominator, that is, from what you divide by. This article covers the formula first, then how the denominator is chosen, and finally the order in which improvements are attempted.

What a conversion means: action, session and user

When asked what a conversion means, the answer "a purchase" is incomplete. A conversion is any action the business cares about and has defined in a measurable way: a form submission, a click on the call button, a price list download, an appointment request. Google Analytics 4 calls these actions "key events", and reporting is built on that definition. The definition is in your hands; so the first job is to put the action that counts as a conversion in writing.

Once the action is defined, the second decision follows: will you divide by sessions or by users? A session-based rate shows how often the decision is made within a single visit. A user-based rate counts multiple visits by the same person once, and is more meaningful in businesses with long decision cycles. In clinics, B2B software and high-priced services, the decision is spread over several days, so a session-based rate looks lower than reality.

Calculating conversion rate: the formula and the denominator

The conversion rate formula is this: conversions divided by the denominator, times one hundred. The denominator can be chosen in three ways. First, total sessions: this gives a rough health indicator for the site as a whole. Second, total users: this counts repeat visits once. Third, relevant traffic: only those who saw the page or channel you want to measure. The third is the most useful in most decisions, because dividing a contact form's rate by visitors who saw the home page and left measures the traffic, not the form.

A simple example shows how much the choice of denominator changes the result. Say a month brings 10 thousand sessions site-wide, 900 sessions on the pricing page and 45 form submissions. Site-wide the rate is 0.45 per cent; on a pricing-page basis it is 5 per cent. Both are correct, but they answer different questions: the first is "how well is the whole site working", the second is "how well does the pricing page carry the decision". If you do not write down which one you put in the report, the same table will be misread two months later.

A third trap is the length of the period. Over short periods the rate swings sharply because the numbers are small: on a site with 12 conversions a week, a difference of two conversions moves the rate substantially in appearance. So the rate should be tracked not weekly but over rolling windows of at least four weeks. We explain the measurement setup and the order in which the first reports should be read step by step in our GA4 article.

Measurement causes that make the rate look low

Before trying to improve a rate, you have to separate whether the low figure is real or comes from measurement. The four causes we see most often in practice are these. First, double counting: the same form is measured both by a button click and by a thank-you page, conversions inflate, and the rate drifts upward. Second, undercounting: conversions that leave the site, such as WhatsApp or phone, are never measured, and the rate drifts downward.

Third, bots and internal traffic: unfiltered visits by your own team inflate the denominator. Fourth, channel mix: if the share of paid traffic rose within a month, the average rate falls even though nothing on the site changed, because visitors with different intent enter the same denominator. For this reason the rate is always read together with a channel breakdown. We go into how to build attribution for contacts that leave the site in our attribution chain article.

The fix order: measurement first, then the page, testing last

Most of the changes proposed for lifting a conversion rate revolve around button colour and headline copy. Those changes belong last, because the larger losses are closed first. The order we apply is this: verify the measurement, clarify the question the page carries, shorten the form, fix speed and mobile layout, and only then test copy and design variants.

Clarifying the question the page carries is the step with the most return. A pricing page answers "how much"; a service page answers "is this right for me". When a page tries to answer both at once, it answers both weakly. We break down which sections a conversion page consists of and which objection each section handles in our landing page article.

Shortening the form comes second and has a measurable cost: fewer fields can mean more requests but less qualified ones. So a form change is assessed not by conversion count alone but together with the count of qualified requests. Speed and mobile layout come third; on a small screen, a call button falling below the first viewport produces more loss than a copy change.

Testing comes last because testing demands sufficient traffic. On a page taking a handful of conversions a week, comparing two variants cannot separate the difference from noise. In such cases dropping the test and looking at qualitative data gives faster answers: session recordings, heatmaps and form abandonment points. We explain the setup of these tools and which question each one answers in our Microsoft Clarity article. We define the scope of conversion work as a service on our CRO and UX page.

The improvement work itself should also be logged. If the date of every change, the page it was made on and its expected effect are noted in a single line, it becomes possible two months later to discuss why the rate changed. Without that log, the team cannot separate a page change from an advertising budget change made in the same period, and the improvement is credited to the wrong work. The most useful format in practice is a one-page change log: date, page, change, expectation, and the observation four weeks later.

How to put the rate in a report

Rather than a single conversion rate, three lines read more accurately: the rate on the relevant page, the channel breakdown, and the length of the period. Next to it the absolute number of conversions should be written, because the rate can rise while the number falls; that happens often when traffic declines. We describe the scope in which we set up the measurement side on our analytics and data page.

One final note: conversion rate is an indicator, not a target. The target is qualified demand; the rate shows where along the path to that target the loss occurs. Teams that turn the rate into a target tend to improve the number by shrinking the denominator, and after a while they produce a report that looks better on less traffic. That is why the rate should always be read alongside the absolute number.

Sources

• Google Analytics Help — About key events: defining and reporting the actions that count as conversions in GA4.
• Google Search Central — Do you need an SEO?: official guidance on how to question measurement and outcome claims.