Funnel analysis: how to find where you lose people

Updated 2026-11-02

Funnel analysis measures how many people complete each step on the way to a goal, and how many drop off between steps. It shows you where you're losing people, so you can fix the step that matters most.

Say 1,000 people visit your pricing page and 20 start paying. That's a 2% conversion rate, and on its own it tells you very little. A funnel shows whether those people left at signup, in onboarding or at checkout. Each of those needs a different fix.

What is a funnel?

A funnel is an ordered list of steps that people take toward a goal. Each step is a page view or an action. A common SaaS funnel looks like this:

1. Visit the pricing page
2. Start signup
3. Complete signup
4. Create a first project
5. Start paying

Funnels narrow at every step. Some drop-off is normal. Your job is to find the step that loses more people than it should.

How to do a funnel analysis

Step 1: Pick one goal

Choose one outcome. Signups, paid conversions and activation are the usual picks. If you try to study everything at once, you'll fix nothing.

Step 2: Map the steps

Write down the path you expect. Keep it short, because fewer steps are easier to act on. Four to six is plenty. Use steps people really must pass through. Optional pages only muddy the numbers.

Step 3: Make sure the steps are tracked

Pages are usually tracked by default. Actions such as signup_completed need a custom event. Check that each step fires reliably before you trust the report.

Step 4: Set your date range and window

Use a date range with enough traffic. Then think about the conversion window, which is how long someone has to finish the funnel. A 7-day window fits many signup flows. A slow enterprise purchase would need much longer.

Step 5: Read the drop-offs

Look at two figures for each step. Step conversion is the share who moved on from the previous step. Overall conversion is the share of the original group who got this far.

Step 6: Segment the funnel

Averages hide things. Split the funnel by device, channel, or new and returning visitors. The problem may sit in one group.

Step 7: Find out why, then test a fix

A funnel shows where. It doesn't show why. Use recordings, heatmaps and user conversations to form a theory. Change one thing. Measure again.

A worked example

Here's a made-up example. A SaaS company looks at last month's funnel, starting with visitors to its pricing page:

The largest drop in raw numbers is pricing to signup, with 2,800 people gone. It's tempting to start there, but pricing pages always lose lots of people, because many are only comparing options. A 30% rate might be fine.

Look at step four. Only 40% of people who sign up create a first project. Those people were interested enough to make an account, and then they stopped.

Now do the math on each fix. Assume the later steps hold steady:

• Raise pricing to signup from 30% to 33%. That's 1,320 signup visits and about 130 paid. You gain about 12 paying customers.
• Raise signup to first project from 40% to 50%. That's 420 projects and about 147 paid. You gain about 29 paying customers.

The second fix is worth more than twice the first, and it's probably easier. So the team watches ten session recordings of new accounts. They see people land on an empty dashboard with no clear next step. They add a sample project and a single prompt. Next month they run the funnel again and compare.

How to segment this funnel

The team then splits the funnel by channel:

The overall 40% was hiding a split. Paid social signups rarely start a project, which suggests the ads are attracting the wrong people or making the wrong promise. So onboarding isn't the only thing to fix. The ad message needs work too.

This is where a funnel and revenue attribution work well together. One shows behavior, and the other shows which channels pay.

Common funnel analysis mistakes

Fixing the biggest drop by default. The biggest drop isn't always the best opportunity. Multiply through to the end of the funnel, as we did above.

Too many steps. A 12-step funnel creates noise. Merge steps until each one is a real decision.

Too little data. With 40 people in a funnel, one person changes a step by 2.5 points. Wait for more data, or read the sessions by hand.

Ignoring order. A proper funnel counts people who did the steps in order. Comparing page totals doesn't do that.

Reading it as a why. A drop-off tells you where to look. Don't assume the cause. Check it.

Changing several things at once. You won't know what worked. An A/B test gives you a cleaner answer.

Funnel analysis versus flows

A funnel tests a path you've defined. A flow chart shows the paths people really took. Use a flow when you don't know the route yet. You may discover that a lot of buyers read a comparison page before pricing. Then you can add that as a funnel step.

Tools for funnel analysis

Most analytics products include funnels, from GA4 to product tools. Look for these features:

• Steps based on both pages and events
• Order enforced within a time window
• Segments by source and device
• A path from a drop-off to recordings or heatmaps

Funnels in Headstream support up to 6 steps. Each one can be a page, a page pattern ending in *, or an event name. Visitors must reach them in order within 7 days. Sankey flows show what people did in practice, and heatmaps and recordings show what happened on the page. You can also save any funnel to a dashboard.

Once you know where people drop off, the next question is which marketing sends the people who stay. Our guide to marketing attribution covers that.