Anyone who has sat through a marketing course has seen the same drawing: an inverted triangle, wide at the top and narrow at the bottom, labeled awareness, consideration, decision. That drawing is about as honest as a subway map: useful for getting your bearings, useless for understanding where the actual tracks run.
The concept is not the problem. The problem is what happens after someone draws that pipe: they build campaigns as if people slid from top to bottom in a single file line, never going back up. Almost nobody moves that way. A person reads an article, disappears for two weeks, comes back through a different ad, asks a colleague, and only then fills out a form. Google's consumer insights team called this the "messy middle" in a 2020 study: between trigger and purchase, people loop through exploration and evaluation as many times as they need to decide. The real funnel has back-and-forth. The drawing does not.
Why "funnel" is the metaphor that misleads
A pipe has one defining property: whatever goes in at the top comes out at the bottom, slower but in the same quantity. In marketing that is almost never true. Most people who enter at the top do not exit at all. They stall, and stay stalled for months. In B2B, Forrester found that 86% of purchases stalled during the buying process in 2024.
That changes everything worth measuring. Think in terms of a pipe, and you watch how many enter. Think in terms of a path with stops, and you watch where people stall and for how long, because that is where the real work is. In our guide on franchise development we described a case where the bottleneck was not traffic at the top: it was a page that failed to answer the questions of people who had already made it to the bottom. Same principle, different industry.
People do not slide down a pipe. They walk a path, and sometimes they turn back.
A funnel is not drawn. You observe it, then build around what you saw.
Which parts of a funnel actually matter?
Once you drop the drawing, three pieces remain, and they need to exist in the opposite order of priority from the one almost everyone picks.
The page that answers real objections. This gets built before the campaigns, not after. Send traffic to a generic page and every dollar spent on ads is only paying to discover that the page does not convert, something you could have learned for free.
Content that reaches people already searching. Not people scrolling by chance: people who already have the question in their head and are typing it into Google or asking a model. This is almost always long-tail, and it converts better than broad campaigns because it arrives at the right moment.
A follow-up sequence for people not ready today. Most people who arrive do not buy right away. LinkedIn's B2B Institute calls it the 95-5 rule: 95% of your potential buyers are not ready to buy today, but will be at some point. Without a sequence that keeps them warm — email, remarketing, new content — that majority is lost for good, and it is the most expensive part to win back later.
- NOIMap the real objections — talk to people who bought and people who said no, and write down what they actually said
- AIProduce the intercept content — articles, pages, ad variants, built around the real questions gathered above
- NOIThe page that closes — written to answer, not to impress
- AIRouting and nurture — anyone who does not convert enters an automated sequence instead of disappearing
- NOIWeekly read of the numbers — what stalled, what unblocked, what moves next week
Where customers actually drop off
| Where people usually look | Where the drop-off really is |
|---|---|
| Traffic at the top | The page that does not answer objections |
| Number of leads | The quality of the filter after the form |
| Cost per click | The time between first contact and decision |
| Email open rate | What happens to people who have not opened anything yet |
The mistake we see most often is putting all the budget into the first column and none into the second. A well-run campaign sending traffic to a weak page is not a funnel that leaks a little: it is a funnel that throws away almost everything it brings in. Fix the page first, then turn on spend. Do it the other way and you burn budget learning something you could have found out for free, just by looking at the page as a stranger would for five minutes.
There is a subtler mistake too, and we made it ourselves early on: measuring only new leads and ignoring the old ones who never got a second touch. In most companies we work with, the "not now" list is bigger than the active lead list, and nobody ever looks at it again. Recovering them is cheap, because those people already raised their hand once.
How do you tell if a funnel actually works?
Three numbers matter more than everything else combined, and they are not the ones most dashboards show first.
The first is the rate of passage from one stage to the next, not the total at the top. A funnel with fewer entries but a good pass-through rate at every stage always beats a funnel with lots of entries and one bottleneck in the middle.
The second is how long it takes between first contact and decision. If that keeps stretching month after month, something in the middle of the path is holding people longer than it should, even if the final number of sales stays the same. In GA4 both of these numbers live on one screen: a funnel exploration shows how many people continue at each step and, with one toggle, the average time between steps.
The third is what happens to the people who do not convert, because it is the part almost everyone ignores, and over time it outweighs almost anything else. A well-built nurture sequence turns a funnel that loses people into one that simply sends them forward later, once they are ready.
AI agents help most with this last part: they read the data every week, flag where a stage got worse, and produce the variants to test. Reading the results and deciding what to do stays with the people who actually know the customer, which is why we cover it in our guide to AI agents for marketing.
Frequently asked questions
What exactly is a marketing funnel?
A funnel is the path someone takes from noticing a problem to buying the solution from you. That path is not linear: it has stalls, returns, and side exits, which is why the old pipe image only tells part of the story.
How many stages should a funnel have?
There is no fixed number. What matters is three functions that always need to be present — reach, convert, and keep warm those not ready yet — not the number of boxes on a diagram. A three-stage funnel that is properly built beats a ten-stage one that is drawn but never measured.
Where do I start when building a funnel from scratch?
With the page that has to close the deal, not with traffic. If it does not answer the real objections of the people who land there, every dollar spent driving traffic to it gets burned for nothing. Turn on spend once the page already converts something.
Why does my funnel bring leads but very few sales?
Almost always because qualification and nurture are missing: leads come in, nobody filters them, and the ones not ready today never get a useful second touch. The traffic is working. What happens after the first form fill is not.
Does it make sense to build a separate funnel for every channel?
No, and trying usually makes things worse. One single funnel with several entry points works better: each channel brings people into the same path, and you compare the traffic coming in, not the funnel itself, which stays one.
Getting traffic but not sales?
Seven questions, and we tell you where customers are actually dropping off. Founders answer within 48 hours. If we are not the right fit, we say so.
ApplyKeep reading
Founder of As Growth, an AI-first growth partner. He runs growth and strategy across partner businesses: OroEtic, Giojello, Equity Plus and others. Apply if you want to work with us.