How to Segment a Funnel:Source, Device, New vs Returning

To segment a funnel, split the same steps by who arrived: traffic source first, device second, then new versus returning. Compare each group's rate at every step. Your blended rate averages audiences that behave completely differently, and it hides the two findings that matter most.

How to segment a funnel before you optimise anything

Your overall conversion rate is a lie. A polite one, but a lie.

It's an average. Some of your visitors came from a post that attracted exactly the right people. Some came from a link that attracted people who were never going to buy. Some are on a phone where your form is a nightmare. The blended number mixes them all into one figure and tells you nothing about any of them.

So you set a goal against it. You try to improve it. And you might be working on the wrong half of your audience.

Segmenting fixes that. To segment a funnel means taking the same funnel and asking the same question of each group separately. It's the cheapest analysis you can do and it routinely changes what you were about to do next.

The two findings an average hides

An overall rate buries two things. Both are worth money.

The segment that converts far better than the rest. This tells you who to go and get more of. It's the closest thing a solo operator has to a targeting brief. It shows where the right people come from and what they do differently.

The segment that converts terribly. It may be dragging your whole number down. And it's often the cheapest thing to fix or simply stop doing. Not every audience deserves your evenings.

Here's the part that stings. The terrible segment is sometimes your favourite channel. The one you've been proud of. It sends plenty of people. They just aren't buyers.

Free guide · 20 pagesStop guessing where it leaks.The Funnel Playbook: where you are losing people and how to find out.Get the free guide →

Which cuts to try first when you segment a funnel

Don't slice everything. Pick the cuts that could change a decision.

The first cut is usually source. The second is usually device. Both routinely change what you would have done. Do those two before you invent a fancier slice.

A note on that third row. New versus returning matters because the same page can behave differently for someone seeing it for the first time and someone coming back to decide. If the returning group converts far better, the job is to bring people back, not to squeeze more out of the first visit.

Source plus funnel beats either alone

Attribution tells you which channel produced a customer. The funnel tells you what happens to the people that channel sends. Together they're far more powerful than either one.

Why? A channel with poor conversion might be a bad channel. Or it might be sending the wrong people, because of how you describe the offer there. You can't tell those apart with only one view.

Judge channels on volume alone and you'll reliably fund whichever source produces the most people who were never going to buy.

That is a particularly easy mistake to make when you're alone. The busiest channel feels like the best one. It's loud, it's flattering and it fills your charts. Segmenting by source shows what happens after the click, which is the only part that pays you.

If a source converts badly across every step, that's a sign of mismatch rather than a page problem. The fix is upstream, in what you say or who you target, and not on the funnel step itself. That's covered in how to fix the biggest funnel drop off.

What to do with each result

A split is only useful if it changes what you do on Monday. So decide the response before you look.

A segment far above the rest. Do more of whatever brings them. Say you sell a productised audit, and people from one community move through every step while people from a general feed barely move. You don't need a better page. You need to spend your evenings where those people are, and describe the offer the way they describe their problem.

A segment far below the rest. Ask two questions. Is it a mismatch, meaning these people were never right for the offer? Or is it a broken experience for them, such as a step that fails on a phone? The first means stop or change the message. The second means a fix, and it's often a cheap one.

No difference at all. That's an answer too. It tells you the leak isn't about who arrives, so the problem is in the steps themselves. Go back and rank the losses.

Whatever you find, don't act on three segments at once. Pick the single clearest gap. One change, one written hypothesis, then look again.

Do you have enough volume to split?

Here is the catch that most segmenting advice skips.

Your funnel is probably small. Cut it into four groups and each one may hold too few people to mean anything. Two people leaving from a group of five tells you nothing.

So before you read a split, ask one question: is there enough volume in each segment for the difference to mean something? If not, say so. Don't invent a story about a slice of nine people.

That has three practical consequences.

• Split fewer ways. Source or device, not source by device by country.
• Use big gaps. A huge difference between two groups is worth acting on. A tiny one is noise.
• Treat small groups as questions. A pattern in a thin slice is a reason to look closer, not proof.

The same honesty applies when you judge a change. Your funnel almost certainly can't detect small improvements. Calling something a reasoned judgment is better than pretending it was an experiment.

Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →

Segment before you fix, not after

Quick version first. If you skip this, you risk fixing the average instead of the problem.

The order matters. Segment first, then optimise.

Imagine you find a step losing the most people. You rebuild it, and the rate barely moves. What happened? Maybe the loss was one audience. Maybe the people leaving were all from one source, and the rest of your visitors were fine all along. You rebuilt a step that wasn't broken for anyone you wanted.

Checking the segments takes minutes. It's the line on the improvement checklist that reads: "I have checked the segments, in case the problem is one audience rather than the step." It sits before diagnosis for a reason.

When you've located your biggest loss, how to find where you are losing customers shows how to rank it, and the segment view is the second pass on top of that ranking.

Two traps when you read a split

First trap: a vivid story. One group looks terrible, so you invent a reason and rebuild around it. The numbers found a location. They did not supply a cause. Watch a few of those people go through the step, or ask some who left, before you decide why.

Second trap: a missing join. If your tracking can't tell that the anonymous visitor and the known customer are the same person, one segment can look like it falls off a cliff at signup when really your tooling lost the thread. Walk a real journey yourself and confirm each segment is recorded correctly before you trust a gap.

Free guide · 20 pagesStop guessing where it leaks.The Funnel Playbook: where you are losing people and how to find out.Get the free guide →

Set it up once, in the definition

Don't leave segmentation for later. Decide it when you define the funnel.

Write down which segments you'll look at first, and that traffic source is one. Check you'll have enough volume per segment. Note what your blended rate hides before you start quoting it. Those are three lines on the definition checklist, and they're cheap to write now and annoying to retrofit.

The reasoning behind that checklist is in how to define a funnel for a small business.

Your next step is small: segment a funnel, the one you're tracking. Split it by source. Write down which source moves furthest and which moves least. Then ask the awkward question: where are you spending your evenings?

Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →

Frequently asked questions

What should I segment a funnel by first?

Traffic source first, then device. Both routinely change what you would have done. After those, new versus returning visitors is a useful third cut. Resist slicing further until the first cuts have changed a decision.

Why segment a funnel instead of trusting my overall conversion rate?

It averages audiences that behave very differently. It hides the segment that converts far better, which shows who to find more of, and the segment that converts badly, which may be dragging the whole number down.

How much traffic do I need to segment?

Enough that each group holds a meaningful number of people. If a split leaves very small groups, split fewer ways, trust only large gaps and treat small differences as questions to look into, not answers.

If one source converts badly, should I drop it?

Not automatically. It may be sending the wrong people because of how the offer is described there. Check whether the problem is the channel or the message. If it is a mismatch, fix it upstream or stop spending time on it.

The full method, with the checklists, is in The Funnel Playbook, a free 20-page playbook.