Funnel Leak

Complaining that conversion is bad is nothing new. The problem was that the complaint pointed nowhere. Put a single number like 1.8 percent on a screen and a meeting can say exactly two things about it: it's low, and it should go up. Which part should go up is a question nobody in the room can answer.

The response was correspondingly blunt. Signups flat? Rebuild the homepage. Still flat? Buy more ads. Teams regularly spent three months replacing screens and watched the number sit still, because the place people were actually leaving was the screen after the homepage, and the homepage was simply the easiest thing to rebuild.

Funnel leak is the idea of giving the problem an address. Break that one conversion rate into steps and the loss turns out to be anything but evenly spread — usually one or two steps lose more than all the others combined. Once that step has a name, the meeting moves from "conversion is low" to "three out of four people leave on the signup screen."

You find one by calculating the pass rate of each step, laying those rates side by side, and picking the step that is unusually low compared with its neighbors.

Take Pagedeck's free signup path: 40,000 homepage visits, 5,000 people reaching the signup screen, 1,200 completed signups, 700 first documents created. Step by step, the pass rates are 12.5 percent, 24 percent and 58 percent.

The leak is not the first step, even though 12.5 percent is the lowest rate on the list. Most visitors arrive with no intent to buy, so that step is supposed to be low. The leak is the second one. When three out of four people who reached the signup screen with an account in mind walk away, something on that screen needs explaining. The benchmark is never the absolute number; it's what that particular step should produce, taken from the same step last quarter or from a different acquisition channel today.

Rates alone tilt the judgment, so count the people lost as well. In the path above, the first step loses 35,000 people and the third loses 500. Lifting the third step from 58 to 70 percent adds 140 people; lifting the second from 24 to 30 percent adds 300 and pushes every later step up with it. That's why teams keep pass rate and absolute loss in the same table before choosing.

One thing remains after the step is identified. A leak tells you where, never what. Open the step and the individual obstacles you find there — the ones that make people hesitate to continue — are friction. A single leak usually holds several at once. Phone verification, a credit card field and a three-second load on the same screen means removing any one of them may not move the number at all.

The most common home for the term is the signup flow, the sequence of screens a person crosses to create an account. It's the narrow neck every user passes through, so a single percentage point there lifts the headcount in every step behind it. Three screens with nine input fields and one screen with two fields lead to the same product, but not with the same number of arrivals.

The second is anywhere money changes hands: the upgrade screen, the payment details screen, the jump to a team plan. Fewer people pass through, but each one is worth more, so a small change in rate shows up directly in revenue.

The third is the first week inside the product. Who signed up and never created a document, who created one and never invited a colleague — the same counting applies. Leaks here are less often a screen problem than a sign that the reason to use the product never landed.

A leak usually turns into an experiment. When the team runs an A/B test, splitting traffic randomly between two versions and comparing results, this is what decides which screen gets tested. It also shapes quarterly goals. Raising conversion rate from 1.8 to 2.5 percent is a wish; raising signup screen pass-through from 24 to 32 percent already names who does what.

"The biggest leak is the payment screen. Five points there changes the monthly number."

A proposal to work on the payment screen, argued from the size of the loss rather than the rate.

"It's a leak, but we don't know the cause yet, so this week we're watching session recordings."

The location is known and the obstacle isn't, so the team observes before changing anything.

"That step isn't a leak. It's supposed to drop that much."

A reminder that a low rate on its own isn't a problem until you compare it with what the step should produce.

At Pagedeck, Ravi once pushed visits from 30,000 to 40,000 with extra ad spend. Completed signups went from 1,200 to 1,210. Mia pulled the step rates and found that signup screen pass-through had fallen from 31 percent to 24 percent — the extra 10,000 people had walked straight out of that step.

The conversation moved from budget to screen. Noa counted the nine input fields out loud, and Theo pointed out that company name and job title could be collected later. All the word did was change the subject of the meeting from ads to one screen, and that swap decided how a quarter of budget got spent.

The most common assumption is that plugging a leak sends everyone who was leaving into the next step. The mental image of water and a hole makes that feel obvious. In practice only some of them continue, and the ones who do can drag the next step's rate down. Lower the signup bar, watch pass-through climb from 24 to 40 percent, and find that the number of people reaching payment hasn't changed — less committed people came in with the rest. So after fixing a leak, teams check the final step's headcount rather than the repaired step alone.

Expecting drop-off to approach zero is the other one. Falling numbers look like pure loss. But every step has a share that was always going to leave: people browsing, people comparing options, people who don't need this yet. Left uncorrected, this belief has teams spending a quarter on a step that cannot be improved while the step that really is below its benchmark goes untouched.