Conversion funnel optimization: fix the step that loses the most people

Conversion funnel optimization is the work of finding the step where most of your visitors give up, fixing that step, and proving the fix with an A/B test. Then you move to the next weakest step and do it again.

Most guides on this topic end in a list of analytics tools. This one ends in test results. Below, we walk through how to map a funnel, how to pick the step to fix first, and what 5 of our published A/B tests showed at 5 different steps: the landing page, the product page, the cart, the lead form and the checkout. 4 won. In 1, visitors preferred the original.

Conversion funnel optimization fixes the 1 step that loses the most visitors

A conversion funnel is the path a visitor takes from arriving on your site to buying or sending a lead. Every step loses some people. Funnel optimization means you stop guessing across the whole site and work on the step where the loss is biggest.

I define CRO as “the process of optimizing a website or app experience to increase the number of desired conversions (purchases, leads, revenue) based on the current traffic and audience.” Funnel work is the same idea, aimed at 1 step at a time. You are not buying more traffic. You are moving more of the visitors you already have through the next door.

Here is why the step view matters. Say 10,000 visitors land on a store in a month. 4,000 reach a product page, 800 add something to the cart, and 200 buy. The overall conversion rate is 2%. That number alone tells you very little. The step rates tell you where to look:

• 40% of visitors reach a product page.
• 20% of those add to cart.
• 25% of carts become orders.

Each of those rates is its own problem, with its own fixes. And a lift at any step carries through to the end. If the add to cart rate in that example rose from 20% to 22%, and every other step held steady, orders would rise from 200 to 220 with no new traffic.

A store funnel has 5 steps and a lead gen funnel has 3

Before you can find the weak step, you need to name the steps. Most sites fit 1 of 2 shapes.

An online store funnel has 5 steps:

1. Landing. The first page a visitor sees. It is often the homepage, a campaign page or a collection page.
2. Product or collection. Where shoppers browse and compare.
3. Cart. Where they check what they picked.
4. Checkout. Where they enter shipping and payment details.
5. Order. The confirmation page that proves the sale happened.

A lead gen funnel has 3 steps:

1. Landing. The page an ad, email or search result sends people to.
2. Form. Where visitors share their details. Some forms run across several pages, and each page is a step of its own.
3. Thank you. The confirmation page that proves the lead came in.

You will also hear “sales funnel.” A sales funnel usually covers the whole buying journey, including ads, emails and sales calls that happen off your site. A conversion funnel is the part you can see in your web analytics: the pages a visitor moves through on your site. That is the part an A/B test can change.

You can find your biggest drop-off in 3 steps

Conversion funnel analysis turns a pile of page views into a clear target. Here is how we do it.

1. Set up the funnel

List the page or event that marks each step, in order. Then count how many visitors reach each one. For a store, the steps might be a landing page view, a product view, an add to cart, a checkout start and a purchase. For a lead form, they might be a landing page view, a form start, each page of the form, and the thank you page. For most sites, we build this in GA4 and spreadsheets. When a funnel is more complex, or has variables like several product types or form paths, I will build it with Claude or a custom GA4 funnel.

Split the funnel by device from the start. Phone and desktop visitors often stall at different steps, and a blended number can hide a problem on either one.

2. Read the step conversion rates

Divide the visitors who reach each step by the visitors who reached the step before it. That gives you a rate for every step. The drop-off is whatever is left over: a step that passes 30% of visitors along loses the other 70%.

Numbers show you where people leave. They do not show you why. Both kinds of research play a role here. Analytics shows you what visitors did, and usability testing can show you the why. Heatmaps can help too, and we cover them in Heatmap optimization trends. This mix of numbers and observation is the core of our Data-Driven Research work.

3. Pick the worst step

When I start with a new site, the first tests depend on what we find in the research phase. We look at the conversion rate of each funnel step and generally attack the step with the highest drop-off. If there is no clear target, we look at the landing pages with the most traffic.

Weigh each drop against how many people it touches. A step that loses 80% of 200 visitors costs you 160 people. A step that loses 50% of 20,000 visitors costs you 10,000. The second step is the better place to start, even though its rate looks healthier.

That choice sets the first tests. It also keeps the program honest, because every test traces back to a number you can watch move. For the full research checklist, see How to do a CRO audit.

70.22% of carts are abandoned, so judge each step against its own past

People often ask what a good funnel conversion rate is. There is no single answer, because each step does a different job.

In ecommerce, I treat the start and the end of the funnel as the most important conversion points. The first step shows how much visitors trust you, and how well your traffic matches the page it lands on. If an ad promises 1 thing and the page shows another, the first step is where you see it.

The middle of the funnel behaves differently. Shoppers there may be in a discovery phase, jumping around and comparing products. Lower step rates in the middle are normal. Deep in the funnel, things should tighten up again: your add to cart to purchase rate should bump back up compared to the middle steps. If it does not, something at the cart or checkout is pushing buyers away.

Outside data shows how much is at stake late in the funnel. Baymard Institute puts the average documented cart abandonment rate at 70.22%, across 50 studies. Among US shoppers who left during checkout, the top reasons were extra costs that were too high (40%), slow delivery (20%) and not trusting the site with their card details (19%).

So use benchmarks as a rough check, not a goal. The most useful comparison is your own funnel: the same step, the same season, the same traffic sources, before and after a change. That is exactly what an A/B test measures. The 5 tests below each worked on a different step.

Links to option pages lifted landing page leads 14.6%

The landing step is the first test of trust. On 1 lead gen site, a landing page sent visitors straight into a sign-up app. For people who already knew what they wanted, that was fast. For everyone else, it was a big ask on the first click.

We tested a softer first step. The variation updated the copy and linked to pages that explained each option first. Those pages then asked for an email.

Leads rose 14.6%. The site had low traffic, so the test ran for 62 days to reach a clear answer. Midway through, the lift was near 20%. It then settled at 14.6%, which is a good reminder to let a test finish instead of calling it on an early peak.

The takeaway: many visitors on a first step are still exploring. Sending them straight into a sign-up flow asked for a decision they were not ready to make. Giving them a page that answered their questions first earned more of them in the end.

Price inside the add to cart button lifted product page conversion 13.9%

The product page sits in the middle of the funnel, where shoppers compare. On 1 store, the add to cart button on desktop had low contrast. It blended into the page, and the price sat apart from it.

We tested a yellow add to cart button with the price inside it. The button stood out, and it answered “how much?” at the moment of clicking.

Conversion rose 13.9% over 24 days, and revenue per visitor rose 16.9%. The change was implemented, and the same idea was later taken to other key pages on the site.

The takeaway: shoppers in the middle of the funnel are scanning. A button they can find at a glance, carrying the 1 fact they need, made the next step easier to take.

The cart step showed a related lesson. In the View Cart button test, we made View Cart the main button in a store’s mini cart, instead of pushing shoppers straight to checkout. Conversion rose 20.0% and revenue per visitor rose 12.8% over 21 days. Shoppers who had just added an item still wanted to review their picks. Letting them do that moved more of them to an order than rushing them did.

A trust seal and fewer distractions lifted lead form conversion 11.6%

In a lead gen funnel, the form is where most visitors stall. On 1 site, an earlier test had added a trust seal to the form. The lift was too small to call, but it pointed in the right direction.

We built on that result. The new variation put the seal on the form’s first page. It also removed all of the content below the form once a visitor started filling it in, so there was less to pull them away.

Form conversion rose 11.6% over 23 days, and lead value per visitor rose 11.0%. The change was implemented. Because the seal and the cleaner page went live together, we cannot split the credit between the 2 changes. For this step, the combined win was what mattered.

The takeaway: once a visitor starts a form, everything else on the page is a chance to leave. Trust matters most at the moment they share their details, so that is where the seal belongs.

Visitors preferred a checkout with no time limit by 2.7%

Checkout is the last step before the order, and visitors who reach it have mostly decided. On 1 desktop checkout, we tested whether a deadline would push more of them to finish. Variation A added a note that the quoted total would expire in 5 minutes. Variation B showed a live 5 minute countdown.

Visitors preferred the checkout with no time limit. In the 5 minute price timer test, the original out-converted the timers by 2.7% over 12 days, and we paused the test. At this step, a clock added pressure, not motivation. It worked against the trust a buyer needs before they pay. That fits the outside data: 19% of shoppers who left a checkout said they did not trust the site with their card details. Anything that makes the last step feel less safe can cost orders.

Urgency is not always wrong at checkout. On a different store’s mobile checkout, a plain timer showing how long the cart was reserved lifted conversion 8.0%. The difference is context. A reserved cart helps the shopper. A price that expires threatens them.

Find the step that loses the most people, prove a fix with a test, then move to the next step.

Funnel optimization runs in 5 steps, then starts again

The 5 tests above came from the same loop. Here is how to run it on your own funnel.

1. Map your steps by device. Name every step, count the visitors who reach each one, and split the counts by phone and desktop.
2. Pick the step with the biggest drop. Compare each step to its own past, not to an industry average. If no step stands out, start with your landing pages that get the most traffic.
3. Write a clear hypothesis. Every test we run starts with the same shape: “By [making this change], we will [cause this effect], resulting in [this number moving].” If you cannot fill in all 3 parts, the idea is not ready.
4. Run the test long enough. We like 2 weeks of data before calling a test, and low-traffic steps can need much longer, as the 62 day landing page test showed. Judge each result on conversion and on revenue or lead value per visitor.
5. Implement or pause, then look up and down the funnel. If we see a winning signal at 1 step, we check whether it holds further up and down the funnel. It often shapes paid ad language and images too.

Run a Shopify store? Shopify conversion rate optimization: what our tests show covers tests built for Shopify funnels.

Fixing 1 step at a time can feel slow. But each win compounds with the next. A better first step sends more people to the product page, a better product page sends more to the cart, and every gain applies to the traffic you already have, every month after you implement it.

Frequently asked questions

What is conversion funnel optimization?

Conversion funnel optimization is the process of finding the step where most visitors leave your site, fixing that step, and proving the fix with an A/B test. Then you move to the next weakest step. It gets more sales or leads from the traffic you already have.

What is a conversion funnel in web analytics?

It is the set of steps a visitor takes on your site before converting, measured in a tool like GA4. A store funnel usually runs from landing page to product, cart, checkout and order. A lead gen funnel runs from landing page to form to thank you page. The report shows how many visitors reach each step.

What is a good funnel conversion rate?

There is no single good rate, because each step does a different job. Compare each step to its own past results. In ecommerce, the first and last steps matter most, and your add to cart to purchase rate should be higher than your middle steps. For context, Baymard Institute puts average cart abandonment at 70.22%.

How do you build a multi-step conversion funnel?

Name each step in order, then mark the page or event that shows a visitor reached it. Count visitors at each step, split by device, and divide each count by the step before it. That gives you a conversion rate for every step and shows where the biggest drop-off is.

What is the difference between a conversion funnel and a sales funnel?

A sales funnel covers the whole buying journey, including ads, emails and sales calls that happen off your site. A conversion funnel is the on-site part you can track in web analytics, from the first page a visitor sees to the order or lead. That is the part an A/B test can change.