A conversion funnel is the sequence of steps a visitor takes from first arriving to completing a goal, such as product page, cart, checkout and purchase.
Funnel analysis counts how many people reach each step and how many drop off between steps, so you can see which step loses the most people and decide what to research and test first.
The numbers tell you where people leave. They don't tell you why.
That second question needs research and, ideally, an A/B test.
This guide covers both, with real results from the LiftArchive test database showing what happened when teams changed a funnel step.
What funnel analysis is (and what it isn't)
Funnel analysis takes a goal and breaks it into the ordered steps that lead to it.
For each step, you count the users who reached it, then compare each count with the one before.
The result is a step-by-step conversion rate and its mirror image, the drop-off rate.
A typical ecommerce funnel looks like this: landing or product page, add to cart, start checkout, enter shipping details, enter payment, purchase.
A SaaS funnel might run from pricing page to signup form to account created to first key action.
A lead-gen funnel might be landing page, form start, form submit, qualified lead.
Three things make funnel analysis useful:
- It localizes the problem. A 2% overall conversion rate is one number. A funnel shows whether the loss happens on the product page or in payment.
- It gives you a baseline for each step. You need the step's current rate and traffic to estimate how long a test on that step will take.
- It shows side effects. A change that pushes more people into checkout can lower the checkout completion rate if the extra people are less committed.
What funnel analysis isn't: a diagnosis. A 45% drop between cart and checkout could come from surprise shipping costs, a slow page, a broken button on one browser or simply people using the cart as a wish list.
You find out which with session recordings, surveys, usability tests and heatmaps, then confirm with a test.
Funnels also count steps that aren't the final goal.
Those intermediate actions, like add to cart or form start, are micro-conversions, and they are often the first metric a funnel test moves.
Funnel analytics: tools and how they count
Funnel analytics tools all do the same basic job, but they count users differently.
Those settings change the numbers you see, so check them before you trust a chart.
Google Analytics 4
In GA4, funnels live in Explore, under the funnel exploration technique.
According to Google's documentation, you can define up to 10 steps, and each funnel is either open or closed:
- Closed funnel: users must enter at the first step. Someone who lands directly on the checkout page isn't counted.
- Open funnel: users can enter at any step. Useful when people reach checkout from many places, but it makes the first steps look leakier relative to later ones.
GA4's funnel exploration can also show elapsed time between steps, the top next actions after each step, a trended line-chart view, and breakdowns by a dimension such as device.
One detail in Google's docs is worth knowing: with a breakdown, users are attributed only to the first breakdown value that applies to them.
Our GA4 for CRO guide covers event setup and explorations in more depth.
Product analytics tools
Event-based tools such as Mixpanel and Amplitude build funnels from tracked events rather than pages.
Two settings matter most. The first is the conversion window: Mixpanel's funnel documentation says users get 7 days by default to complete a funnel from the time they perform the step 1 event.
The second is step order: Mixpanel's default "specific order" requires steps in the sequence you defined, while "any order" relaxes that.
Mixpanel also offers exclusion steps (a "did not do" filter between steps) and holding a property constant, which requires a user to keep the same value for an event property across steps.
Holding the product ID constant, for example, stops a funnel from counting someone who viewed product A and bought product B as a conversion.
Settings to check before you compare numbers
| Setting | What it changes | When it misleads |
|---|---|---|
| Open vs closed | Who counts as entering the funnel | Comparing an open funnel in one tool with a closed one in another |
| Conversion window | How long users have to finish | Long purchase cycles (B2B, furniture) look worse with a short window |
| Step order | Whether out-of-order paths count | Users who go back to the product page mid-checkout |
| Counting unit (users vs sessions) | Whether a return visit can complete the funnel | Session funnels undercount people who buy on a second visit |
| Cross-device identity | Whether phone-to-laptop journeys join up | Mobile looks worse when people browse on phones and buy on laptops |
Conversion funnel analysis: a step-by-step method
Conversion funnel analysis turns raw step counts into a short list of things to research and test.
Here is a method that works for most sites.
- Pick one goal and define the steps. Use events that every converting user must fire. Keep it to 4-7 steps; more than that and the counts get thin.
- Pull counts for a stable period. Four weeks smooths out weekday swings. Avoid windows that include a big sale unless you analyze it separately.
- Calculate step conversion and drop-off. Step conversion = users at step / users at previous step. Drop-off = 1 minus step conversion.
- Segment. Split by device, traffic source, new vs returning and country. A step that looks fine overall often hides one broken segment.
- Research the worst gaps. Watch recordings of people who dropped at that step, run an exit survey, test the step on a real phone.
- Write a hypothesis and test it. Measure the step metric and the final goal. Our prioritization guide covers how to rank the candidates.
Worked example
Here is a hypothetical store with 50,000 users viewing a product page in four weeks.
| Step | Users | Step conversion | Drop-off | Share of start |
|---|---|---|---|---|
| Product page view | 50,000 | - | - | 100% |
| Add to cart | 6,000 | 12.0% | 88.0% | 12.0% |
| Begin checkout | 3,300 | 55.0% | 45.0% | 6.6% |
| Shipping details entered | 2,100 | 63.6% | 36.4% | 4.2% |
| Purchase | 1,450 | 69.0% | 31.0% | 2.9% |
The overall rate is 1,450 / 50,000 = 2.9%.
It equals the product of the step rates: 0.12 x 0.55 x 0.636 x 0.69 = 0.029.
That product is the most useful fact in funnel analysis.
Because the steps multiply, a 10% relative improvement at any single step raises purchases by 10%, from 1,450 to about 1,595, whichever step you improve.
The 88% drop at add to cart looks like the biggest problem, but it isn't automatically the best place to work.
What separates the steps is how easy they are to move and how fast you can test them.
The product page has 50,000 users, so a test there reaches a reliable answer quickly.
The shipping step has 2,100, so a test there needs a much larger effect or a much longer run.
Use a sample size estimate before you commit to a step.
When this method goes wrong
- Small counts. A step with 200 users a month will swing by several points from random noise. Don't chase week-to-week changes there.
- Steps that aren't really sequential. If many buyers skip the cart through express checkout, a cart step in a closed funnel undercounts real progress.
- Optimizing a step instead of the goal. Pushing more people into checkout can lower checkout completion and leave revenue flat. Always read the final metric.
The last point shows up in the test record.
When MarketingExperiments tested RegOnline's homepage, a three-step free trial signup that asked for contact details first produced 24.5% fewer completed leads than the two-step version. Adding a step cost conversions.
But the step count alone doesn't decide it.
HomEquity Bank split a one-page reverse mortgage inquiry form into seven steps and saw a 154% higher conversion rate, according to a MarketingSherpa case study.
What matters is what each step asks for and how it feels, which is why the funnel points you to a step and research tells you what to change.
The ecommerce conversion funnel
The ecommerce conversion funnel usually runs: entry page, product page, add to cart, cart, checkout start, shipping, payment, order confirmation.
Most stores lose the majority of visitors before add to cart, then lose a large share again between cart and purchase.
Baymard Institute's cart abandonment research puts the documented average cart abandonment rate at 70.22%, based on 50 studies.
Baymard's survey of US online shoppers also found that about 42% had abandoned a cart because they were just browsing or not ready to buy.
Some of your cart drop-off is normal and not fixable.
Among shoppers who were not just browsing, Baymard lists these as the top reasons for abandoning during checkout:
| Reason (Baymard, US online shoppers) | Share | Funnel step to look at |
|---|---|---|
| Extra costs too high (shipping, tax, fees) | 40% | Product page and cart: show costs earlier |
| Delivery was too slow | 20% | Product page and shipping step |
| Didn't trust the site with card information | 19% | Payment step |
| Site wanted me to create an account | 18% | Checkout start |
| Checkout too long or complicated | 17% | Every checkout step |
| Website had errors or crashed | 17% | Segment by browser and device |
That table is survey data: what shoppers say.
It tells you what to look for, not what will win on your site.
For your own baseline, see our cart abandonment rate benchmarks.
What tests show when you change a funnel step
Removing or merging steps is a common funnel test, and it often wins.
Bonafide Health's cart drawer sent shoppers to a separate cart page before checkout.
A redirect that cut that extra cart-page step raised the checkout rate by 5.6%, in a test published by Visually (vendor case study, sample size not stated).
Britannica's subscription path is an older example. MarketingExperiments reported that removing one page from registration produced 117% more free trials and 130% more revenue.
The revenue number matters: the extra trials turned into money, not just more people at the next step.
Changing what a step shows can work as well as removing it.
ReplaceDirect added an order overview to checkout step 2, along with step-specific benefits, a cleaner layout and fewer form fields.
It reported 25% lower cart abandonment and 12% more sales.
Several changes shipped together, so we can't credit any single one.
Not every shortcut helps. Blend Commerce tested a Shopify express checkout button against payment icons on a travel brand's product page.
The express checkout variant cut mobile conversion by 14% and mobile revenue per visitor by 21%.
Letting people skip the cart isn't always what they want.
The same pattern holds outside ecommerce. DashThis used session recordings to find that the integration buttons in its free-trial onboarding were pale and sometimes below the fold.
Bolder buttons and a reorganized list were among the fixes that came with 50% higher onboarding funnel completion, measured before and after rather than in a split test.
More funnel-step tests from the database:
@oliverkenyon
DRIP Agency
Conversionista
OptiMonkFor a full walkthrough of checkout steps and what to test at each, see our checkout flow guide.
Drop-off rate: how to calculate and read it
Drop-off rate is the share of users who reached one funnel step but didn't reach the next. The formula is:
Drop-off rate = (users at step N - users at step N+1) / users at step N
In the worked example above, 3,300 users began checkout and 2,100 entered shipping details.
The drop-off is (3,300 - 2,100) / 3,300 = 36.4%.
Some reports show a cumulative drop-off from the top of the funnel instead.
By the shipping step, cumulative drop-off is (50,000 - 2,100) / 50,000 = 95.8%.
Both are correct; just don't compare a step drop-off from one report with a cumulative one from another.
Reading drop-off rates well
- Compare against your own history first. A step that jumps from 30% to 40% drop-off after a release points to a bug. That is a faster win than any test.
- Compare segments, not just totals. If mobile drop-off at payment is 50% and desktop is 25%, look at the mobile payment form before anything else.
- Expect high drop-off at the top. Product page to cart drop-off of 85-95% is common because most visitors are browsing. A high number there doesn't mean the page is broken.
- Check the next action. GA4's next action view shows where people went instead. People who leave checkout for the shipping policy page are telling you something different from people who close the tab.
When drop-off rate is the wrong metric
Drop-off rate is a poor guide when steps are optional or reorderable, when the conversion window is shorter than your buying cycle, or when counts are small.
In those cases, track the final goal per visitor and use the funnel only to spot where behavior changed.
It is also the wrong success metric for a test.
A change can lower drop-off at one step by pulling less-committed people forward, then lose them at the next.
Judge tests on purchases, revenue per visitor or qualified leads, and check statistical significance before you call a winner.
Exit rate: what it tells you and what it doesn't
Exit rate is the percentage of a page's views that were the last in a session.
Google's Universal Analytics documentation defined it this way: for all pageviews to the page, exit rate is the percentage that were the last in the session.
Exit rate = exits / views
If your shipping policy page had 10,000 views and 2,500 of them ended the session, its exit rate is 25%.
GA4 doesn't give you an exit rate metric by default.
Its metric definitions include Exits ("the number of times that the last event recorded for a session occurred on a page or screen") and Views.
Add both to an exploration and divide them yourself.
Exit rate vs drop-off rate vs bounce rate
| Metric | Denominator | Question it answers |
|---|---|---|
| Exit rate | All views of the page | How often is this page the last one people see? |
| Drop-off rate | Users who reached a funnel step | How many didn't reach the next step of this goal? |
| Bounce rate (GA4) | All sessions | What share of sessions weren't engaged? |
The differences matter. A user can leave the checkout funnel without exiting the site, for example by going back to browse products.
That counts as a drop-off but not an exit.
In GA4, bounce rate is now the inverse of engagement rate, not the old single-page measure; our bounce rate guide explains the change.
When exit rate misleads
- Pages that should end sessions. An order confirmation page or a contact page with a phone number will have a high exit rate because it did its job.
- Pages ranked without context. Sorting all pages by exit rate puts low-traffic pages at the top. Filter to pages with enough views to matter.
- Pages inside a funnel. For a checkout step, the funnel's drop-off rate is a better measure, because it counts people who left the goal, not just the site.
Exit rate is most useful on pages that should lead somewhere: product, cart and pricing pages.
A high exit rate there is a prompt to research.
OptiMonk's case study with Gelpro Australia is one response to cart exits: an exit-intent popup reminding visitors of the items in their cart came with a 6% higher ecommerce conversion rate, measured before and after rather than in a split test.
A popup treats the symptom. Finding out why people leave the cart is the fuller fix.
Funnel analysis tells you where to look. The tests in our database show what happened when teams acted on it: removing a step often helped, adding one sometimes hurt, and the result always depended on what the step asked of people.
Browse more checkout, cart and form tests in the database before you design your own.