What is a conversion funnel?

The sequence of steps a visitor moves through toward the site's goal — e.g. visit → product view → cart → order. Funnel analysis measures how many people fall out between each pair of steps.

Where is funnel analysis in GA4?

In the Explore section, pick the Funnel exploration template and set your events as the steps — from landing on the site through to the conversion event.

What drop-off between steps is normal?

There's no universal benchmark — it depends on the industry and the step. More useful than someone else's numbers is comparing your own steps against each other (which loses most) and over time (which is getting worse).

What if the events for the funnel steps are missing in GA4?

Add them. Without events for the key steps (form view, submission, add to cart…) you can't build the funnel at all. Basic events can be added in GTM without touching the code.

Funnel analysis in GA4: how to find where you lose customers

You know how many people came to the site. You know how many bought or enquired. What happened to everyone in between — where exactly they got lost and why — is a question most companies answer with a hunch.

GA4 has a tool for exactly that question: funnel analysis. This guide shows how to build one, how to read it, and how to turn it into a decision rather than a chart.

What you need before you start

1. A defined goal. An order, a submitted enquiry, a signup — one main conversion that the funnel ends on.

2. Events for the key steps. GA4 measures everything as events. It collects basics automatically (page_view, scroll, click); e-commerce events (add_to_cart, begin_checkout, purchase) come from the site if they’re implemented. For steps with no events — say viewing and submitting an enquiry form — you have to add them, usually through Google Tag Manager.

This is where most analyses fall over: you can’t build a funnel out of steps nobody measures. If events are missing, implement them and come back in a few weeks — with data.

Step 1: Build the funnel in Explore

GA4 → Explore → the Funnel exploration template.

Set the steps to the real customer journey. For an e-commerce site, typically:

session_start — visit
view_item — product view
add_to_cart — added to cart
begin_checkout — checkout started
purchase — purchase

For a B2B site with an enquiry form:

session_start — visit
offer page view (page_view filtered by URL)
form page view
form submission (custom event)

Two practical settings: an open funnel counts people who entered mid-journey — better for a first overview. And leave the step setting on the default “is indirectly followed by”; requiring steps to follow directly only makes sense in detailed analyses.

Step 2: Find the biggest hole

The funnel gives you two numbers between every pair of steps: how many continued and how many dropped off. The first read is simple: which transition loses the most people in absolute terms?

Watch the percentage trap. A 90 % drop-off on a step a hundred people a month pass through is a smaller problem than a 40 % drop-off on a step thousands pass through. Multiply: people in the step × drop-off = how many customers you leave there each month. That ranking is your priority list.

The second read: comparison over time. Switch the range to the previous month or quarter. A step that’s deteriorating is more urgent than one that’s been consistently weak — something changed, and it’s worth finding out what.

Step 3: Break the funnel into segments

An aggregate funnel averages different groups into one number, and averages lie. Three breakdowns reveal the most:

Device (mobile vs. desktop). The single most common finding: a step that works on desktop and bleeds on mobile. A broken form, overlapping elements, an unreachable button.

Traffic source. Visitors from ads behave differently from organic ones. If paid traffic falls out at the very first step, the problem may not be the site — it may be a mismatch between the ad and the page it leads to.

New vs. returning. Returning visitors convert better almost always. What’s interesting is by how much — an extreme gap suggests the site can’t convince anybody on a first visit.

In Explore you add these through “Breakdown” — drag a dimension (Device category, Session source/medium) into the field.

Step 4: From “where” to “why”

The GA4 funnel tells you where customers drop off. It doesn’t tell you why — and this is where analysis often stops early and turns into speculation.

Two routes to “why”:

Look at behaviour on the problem page. Session recordings or heatmaps show what people actually do there: where they stop, what they click on in vain, when they leave. Watching five real users surfaces the barrier faster than another month of data.

Walk the step as a new visitor. Open the problem page in a private window on a phone and try to complete the step. A good share of findings need no special tooling — just the willingness to see your own site without operational blindness.

Step 5: Turn the finding into a hypothesis and a test

A finding (“twice as many people drop out of the form on mobile as on desktop”) isn’t a solution yet. The solution is a hypothesis: because the mobile form requires nine fields across three screens, we believe cutting it to four will raise completion.

An A/B test then verifies it — and only the result tells you whether you actually plugged the leak or just moved it.

This cycle — funnel → segment → why → hypothesis → test — isn’t a one-off. The funnel changes with every campaign, season and site edit. An analysis six months old describes a site that no longer exists. Which is why this loop is increasingly watched by software rather than a person — see what an AI CRO agent is.

The takeaway

Funnel analysis in GA4 is the fastest route from “conversions are low” to the specific place responsible. The process: build the funnel from events → find the transition with the biggest absolute loss → break it down by device and source → work out why by looking at real behaviour → write a hypothesis and test it.

The one weakness of that process: somebody has to run it regularly. That’s the loop Kaen runs — it watches your funnel, flags what got worse, and proposes a test with the hypothesis attached. You just approve it.