How to find where website visitors drop out with Google Analytics

You find where visitors drop off by turning the path to a conversion into a short list of steps and measuring how many people make it from each step to the next. For a software company, the steps might be landing on the site, viewing the pricing page, starting the signup form, and finishing it. For a services business, they might be landing on a service page, opening the contact page, and submitting the form. The step where the count falls the most is where you're losing people, and that's where your effort pays off first.

Google Analytics 4 gives you two ways to see this. The quickest is the standard Landing page report, which shows each page people arrived on with its sessions, average engagement time, and key events, the actions you've marked as conversions. A landing page with plenty of sessions and almost no key events is a page that brings people in and fails to move them along. The fuller view is a funnel exploration, which you build under Explore. You name each step, define it by an event or a page view, and Google Analytics shows how many active users reached each step, how many abandoned it, and the completion rate between steps. You can break the funnel down by a dimension such as device category or session source, and you can turn on the next action view to see what people did instead of taking the next step. A path exploration works from the other direction: you set the conversion as the end point and see the pages people visited before it. If you sell online and send ecommerce events, the built-in Checkout journey report already shows the funnel from starting checkout to purchase.

These tools are good at counting, and they have limits worth knowing. A funnel can only use events your site sends. Google Analytics' enhanced measurement records page views and scrolls automatically and can record form starts and submissions, but it doesn't see every form, especially one embedded from another tool, such as a HubSpot or Typeform form. If a step depends on an event that never fires, the funnel shows nearly everyone leaving there, even though they didn't. Explorations also only reach as far back as your property's data retention setting, which is two months unless someone changed it. And none of these reports tells you why people leave. A step with a 70 percent drop could have a confusing form, a slow page on phones, a price that surprises people, or visitors who were never a fit to begin with.

A good result is a specific answer: the step that loses the most people, the group of visitors who account for most of that loss, the most likely reason backed by what's on the page, and the change worth testing first. Kite does that analysis for you. It reads your Google Analytics property, builds the funnel, finds where and with whom the loss happens, and then looks at the page to explain it.

Start by telling Kite what you've noticed:

"We get decent traffic but hardly any demo requests. Can you find where people are dropping out?"

Kite is an AI marketer. It researches your business, builds a growth strategy, and ships the work that brings you new customers. From that research, it already knows what your site sells, which pages lead toward a signup or a purchase, and what a conversion means for your business. So it can propose the funnel steps itself, such as landing page, pricing page, demo form started, and demo form submitted. Confirm the steps or change them, and Kite starts.

You can work with Kite in the Kite web app or in Slack. Connect your Google Analytics account through OAuth so Kite can read your reports. If your account can see more than one property, Kite lists them and confirms which one measures the site you're asking about.

Reading reports is routine work, so Kite starts as soon as the account is connected and reports what it finds in the conversation. It asks for your approval before it publishes any change to your site.

Kite starts by finding out what your Google Analytics property can measure, because a funnel is only as good as the events behind it. It reads the property's settings, its data streams, and its data retention period, which tells Kite how far back an exploration can reach. It reads the dimensions and metrics available on the property, including any custom dimensions your team created. It also lists your key events and runs a report of every event your site sent over the period, with how often each one fired.

That inventory tells Kite which steps it can measure directly. A page view of your pricing page and a form_submit event can each define a step. When a step you care about has no event, such as a signup button click or a form embedded from another tool, Kite tells you which event is missing and how to add it. Kite reads your key events but doesn't mark new ones, so when your conversion event exists but isn't marked, Kite tells you which one to mark.

Kite also checks that data is still arriving. It runs a realtime report to confirm events are coming in, and it reads the annotations on your property, which often record a site launch or a tracking change that would split the data in two.

Kite reads the Landing page report before it builds the funnel, because the first place people leave is often the page they arrived on. It compares sessions, engagement rate, bounce rate, and key events for each landing page over the period, and it ranks the pages by how many visitors they bring in without moving them toward a conversion.

It breaks those numbers down by session source and medium and by device category. A landing page that converts well for visitors from search but poorly for visitors from a paid campaign usually has a mismatch between what the ad promised and what the page says. A page that works on desktop but loses most phone visitors usually has a problem only phone visitors see.

Kite runs a funnel report on your property with the steps you confirmed, each defined by the page view or event that marks it. For each step, it reads how many active users reached it, how many abandoned it, the completion rate to the next step, and the abandonment rate. It compares the current period with the one before so you can see whether a step has always been weak or got worse recently. When a step got worse, Kite runs the funnel as a trend over time to find the day the change started.

Kite also reads the next action for each step, meaning the page or event that people reached instead of the next step. People who leave a pricing page for the docs are behaving very differently from people who leave it for nothing at all, and the next action shows which one you're looking at.

When a step shows almost everyone leaving, Kite checks the event behind it before it calls it a real loss. If the next step depends on an event that barely fires anywhere on the site, the gap is a tracking problem, and Kite reports it as one.

Kite breaks the weakest step down by device category, session source and medium, landing page, and country, and it compares new visitors with returning ones. A funnel report can carry one breakdown dimension and up to four segments at a time, so Kite runs as many reports as the question needs.

This is where the analysis usually narrows. An overall drop of 60 percent at the signup form might turn out to be a 30 percent drop on desktop and an 85 percent drop on phones. Kite ranks the groups by how many visitors each one loses at that step, so you see the loss that matters most to your results, not just the worst-looking percentage on a small group.

Kite looks at the page where the loss happens to find the reason, because Google Analytics can't tell you that part. It renders the page and captures screenshots, so it sees what visitors see, including on a phone-sized screen when the loss is on mobile. It measures the page's speed and technical quality with PageSpeed data. It reads the copy, the form fields, the calls to action, and the price, and compares them with what the traffic source promised, such as the ad or search result that sent people there. It also checks the annotations and the trend for the date a step got worse, and compares the page with its archived versions when they exist.

Kite separates what the data shows from what it infers. A slow page on phones and a mobile-only drop at the same step line up well, and Kite says how confident it is. A form with nine fields and a 70 percent abandonment rate is a strong lead, but it isn't proof until a change is measured.

Kite reports its findings in the conversation where you asked. It opens with the answer: the step that loses the most visitors, how many it loses over the period, and which group of visitors accounts for most of the loss. Then it gives the most likely reason with the evidence behind it, any tracking gaps it found, and the change it recommends testing first. Every figure carries the dates it covers. When the findings are too long for one message, Kite publishes them as a findings page with the funnel chart and tables and links to it from the conversation.

From there, Kite can make the change. On a site Kite hosts, or one it updates through a supported CMS such as Webflow, it can shorten a form, rewrite a headline so it matches the ad that sends traffic to it, move a call to action, or fix what slows the page on phones, and it asks for your approval before publishing. On a site Kite hosts, it can run the change as an A/B test with a traffic split between versions. On a site measured with Google Analytics, it compares the same funnel step before and after the change to show whether more visitors made it through.

If you'd like to keep an eye on your funnel, ask Kite to run the same analysis every week and send the results where you choose. Each run compares the week with the one before, so a step that suddenly gets worse shows up quickly.

When Kite finishes, you know where your site loses visitors, which visitors it loses, and what to change first.