How to Find Where You Are Losing Customers in Your Funnel
How to find where you are losing customers: rank every step by people lost, not percentage, cut by source, and walk the path yourself before you trust a number.
To find where you are losing customers, write your steps in order, count people who passed every earlier step, and multiply each drop-off by the volume entering it. The biggest absolute loss is your leak, not the ugliest percentage. Then verify the numbers by walking the path yourself.
How to find where you are losing customers (and stop guessing)
You already have a theory. Everybody does. The pricing page is too long. The headline is weak. The form is annoying.
Maybe you're right. Probably you're not, and here's why. You know where every button is. You understand your own vocabulary. You have never once arrived at your own homepage with no context and a mild interest.
Your intuition about your own product is systematically wrong. So the job isn't to think harder. It's to measure, in a way that can't flatter you.
This post is the finding part. What to do once you've found the leak is a separate job, covered in how to fix the biggest funnel drop-off. Don't skip ahead to fixing. A fix aimed at the wrong step is the most expensive kind.
Start by getting a number you can trust
Three conditions, or the numbers are decoration.
Steps are states, in order. Arrived, gave an email, booked, paid. If someone can reach step four without step two, the chart lies.
Cohorts, not totals. Each step counts only people who got through every earlier one. If your tool just counts how many people did each thing, the thousand at the top and the hundred at the bottom may be different people. You'll see a funnel shape and read fiction.
A stated window. How long someone has to take the next step before you call them lost. Say it every time you quote a rate.
If you haven't set those up, start with how to define a funnel for a small business. Then come back. There's no point hunting a leak with a broken gauge.
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 →Where you are losing customers: rank by people lost, not percentage
This is where most people go wrong, and the maths is kind of embarrassing.
When you check where you are losing customers, you scan the funnel and spot the step with the ugliest rate. You decide that's the problem. It frequently isn't.
Multiply the drop-off by the volume at each step. Then rank by the result.
| Step | Enter | Convert | Lost | Looks like |
|---|---|---|---|---|
| Starts the signup | Two hundred | Ten per cent | One hundred and eighty | The disaster |
| Reads the offer page, where you are losing customers without noticing | Five thousand | Sixty per cent | Two thousand | "Fine, that's decent" |
Those are the guide's own numbers. A step converting at ten per cent with two hundred people entering loses a hundred and eighty. A step converting at sixty per cent with five thousand entering loses two thousand.
The "decent" step is leaking eleven times the people. The "disaster" is a rounding error in comparison.
Ten per cent of a small number is a smaller prize than a modest gain on a large one. So you rank by humans lost. Not by how bad the percentage looks on a slide.
Cut it by source and by device
One more pass before you decide anything. A single overall rate is an average across audiences that behave completely differently.
The first cut is usually traffic source. The second is usually device. Both routinely change what you would have done.
- One source converts far better than the rest. That tells you who to go and get more of.
- One segment converts terribly. It may be dragging the whole number down, and stopping that work can be the cheapest fix you'll find.
- Sometimes the "leaky step" is really one audience that was never right for you.
Check this before you rebuild anything. Otherwise you'll polish a step that was losing only the wrong people. How to do the split properly, and what volume you need for it to mean something, is in how to segment a funnel.
Walk it yourself
Before you believe a single number, go through the funnel as a stranger would.
Use a device that isn't yours. A browser with no history. Arrive from a real link, not from your bookmarks. Then confirm your own journey appears in the data, with every step in the right order.
Twenty minutes. It catches the errors that otherwise cost you a month of confident decisions:
- The event that fires twice, inflating a step.
- The step that never fires on mobile.
- The redirect that drops everything on the way through.
- The tracking that never ran because the visitor declined consent.
Do it once on desktop and once on a phone. A cliff in your chart might not be customers leaving. It might be your tracking losing the thread. How to set that up so it holds is in how to track a funnel with events.
Look at it for a month before you touch it
You'll want to react on day two. Don't.
You need a baseline and a sense of ordinary variation. Without it, you'll react to a normal Tuesday and call it a trend. A change made before you know the baseline can never be evaluated. And that first month is the only clean read you'll ever get of what the funnel does when nobody is interfering with it.
Write down the starting rate at every step. Note how much it wobbles week to week. Sanity-check it against something you know independently, like the number of actual orders in your payment records.
If the funnel says one hundred people paid and your bank says sixty, the funnel is wrong. Find out why before you do anything else.
Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →What to do with the answer
Do not fix it yet. First write down what you believe is happening and why, as a plain sentence. Then check it against what you see when you watch real people.
You'll often be wrong at this point, and that's fine. Being wrong on paper costs a minute. Being wrong after you've rebuilt a page costs a weekend.
Also decide what you will compare against. If you don't know what the step converts today, to the decimal, you will never know whether your change did anything. Record the number now.
Numbers say where. They rarely say why.
Once you've found the step, resist the urge to explain it from your armchair. The funnel gives you a location and sizes the prize. It almost never gives you the cause.
For the cause, watch a handful of people go through the failing step. Or ask a few who abandoned it. An hour of that produces better hypotheses than a week of staring at percentages.
Use both, though. Numbers alone give you blind optimisation. Anecdotes alone have you chasing one vivid story that doesn't generalise. The numbers find the leak. The people explain it.
The ninety-minute version
If you only have one evening, do this in order.
- Write the five or six states. On paper, in order, ending at the money.
- Pull the counts. Cohort counts, with the window written beside them.
- Multiply and rank. Drop-off times volume, step by step.
- Split the top step by source. Then by device.
- Walk it on your phone. From a link you did not make yourself.
- Write one sentence. "We lose most people between this and that." Put it where you'll see it.
That sentence is the output. Not the dashboard. It tells you where you are losing customers. A vague Sunday-night worry has just become a location, a plausible cause and an obvious next task. That conversion, from a mood into a task, is most of the value of measuring when you work alone.
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 →Two answers that mean you're looking in the wrong place
Sometimes the hunt turns up nothing, and that's information.
Every step converts reasonably, and the totals are still small. You don't have a leak. You have a volume problem, and more optimising is misdirected.
Every step is mediocre and none is catastrophic. Suspect the offer, not the sequence. A uniformly poor funnel is usually a positioning problem wearing a conversion costume.
The funnel is a good enough instrument to tell you when it's the wrong one. Listen to it. The most common failure in measurement work is polishing the thing you can see because you can see it.
Free assessmentHow trapped are you, really?Take the Corporate Suffocation Index, the free solopreneur readiness assessment.Take the assessment →Frequently asked questions
How do I know which step is losing the most customers?
Multiply the drop-off by the number of people entering each step, then rank by the result. The step losing the most people is your biggest leak, even if its percentage looks acceptable. Weight that by how cheap the fix would be.
Why do my funnel numbers not match my sales?
Usually the tool counts events instead of cohorts, an event fires twice, or a step never fires on mobile. Walk the funnel yourself from a different device and compare the result against your payment records.
Do I need a lot of traffic to find the leak?
You need enough to read the steps, and a small funnel can still show a large, obvious gap. What you shouldn't do is treat small differences as proof. Look for big losses, and say honestly when the volume is too thin.
What if no obvious leak shows where you are losing customers?
Then the funnel is probably not your constraint. Either volume is too low, or every step is mediocre because the offer is wrong. Neither is fixed by tweaking a page.
The full method, with the checklists, is in The Funnel Playbook, a free 20-page playbook.
How trapped are you, really?
Take the Corporate Suffocation Index, the free solopreneur readiness assessment.