Where the shape came from
A funnel is not a marketing invention. It is an admission — that most people who hear about you will never buy from you — followed by a decision to be precise about where they go instead. Everything useful about funnels comes from taking that admission seriously. Everything embarrassing about them comes from pretending it away.
The shape is old. In 1898 an advertising man named Elias St. Elmo Lewis sketched a sequence he thought every purchase passed through: attention, interest, desire, action. It became AIDA, then the purchase funnel, and it has survived more than a century of fashion because the observation underneath it is simply true. At every step from “heard of you” to “paid you,” the group gets smaller. Draw that honestly and you draw a funnel. There is no way around the narrowing; there is only the question of how steep it is and where.
What has not aged well is everything built on top of that observation. The funnel got sold as a machine — put strangers in, get customers out, turn the handle faster for more revenue. It got drawn as a smooth cone, as though people slide down it in one uninterrupted motion. It got treated as a thing you buy rather than a thing you measure. And it got an ending, at the moment money changes hands, which is the single most expensive mistake in the whole model.
So this piece is not a defence of the funnel or an argument to replace it with something with a more fashionable shape. It is an attempt to describe what the thing actually does, stage by stage, including the parts most explanations skip.
What a funnel actually is
Start here, because almost every bad funnel begins with the wrong definition. A funnel is a measurement device before it is a marketing asset. Its first job is not to sell anything. Its first job is to tell you which step is losing people. A funnel you cannot read is not a funnel; it is a diagram of one.
That reframing changes what “building a funnel” means. It is not the act of buying software or assembling pages. It is the act of making a decision legible — breaking the distance between a stranger and a customer into steps small enough that you can tell which one is failing. Once you can see that, the marketing work becomes obvious. Before you can see it, every improvement is a guess.
Three things a funnel genuinely is:
- A map of a decision. Not your sales process — the customer’s decision process. These are different, and confusing them is why so many funnels feel like being processed rather than being helped.
- A measurement frame. Stages exist so that a number can be attached to each one. A stage you cannot measure is a story you are telling yourself.
- A sequence of promises. Each stage makes one, and the next stage either keeps it or breaks it. Almost every conversion problem is a broken promise wearing a different costume.
And three things it is not. It is not a landing page — a page is one stage, not the system. It is not a piece of software; tools record a funnel, they do not constitute one. And it is not a guarantee: a funnel does not create demand that was never there. It organises demand that already exists. Pointing a beautiful funnel at people who do not want the thing is the most elaborate way to learn that fact.
A funnel you cannot read is not a funnel. It is a drawing of one.
The six stages, honestly
The classic model has four stages. The honest version has six, and the two extra ones are where most of the money is. Here is each stage described by what the person is doing, not what your marketing is doing — because the person is the only one whose behaviour you are actually trying to understand.
1. Attention — they notice
Someone becomes aware you exist. That is the entire stage, and it is worth almost nothing on its own. Attention is the cheapest thing to buy and the easiest thing to mistake for progress; a spike in traffic with no movement anywhere else is not a good day, it is an expensive one. Your job here is narrow: be seen by people who could plausibly want this, and say something specific enough that the right ones lean in and the wrong ones keep walking. Attention that does not qualify is just noise you paid for.
2. Connection — they recognise something
This is the stage the classic model skips, and skipping it is why so many funnels feel cold. Between noticing you and evaluating you, something has to happen: the person has to feel that you understand their situation. Not that your product is good — that you have seen their problem clearly enough to describe it back to them better than they could.
You cannot argue someone into this. It happens in the first few seconds, in the specificity of your language, in whether you name the thing they actually worry about at 1am or the sanitised version of it. When connection is missing, every downstream number looks fine and nothing converts, because people are reading your page as information rather than as something addressed to them.
3. Clarity — they understand the offer
Now they are paying attention on purpose, and the question changes from “is this for me?” to “what is it, exactly?” This stage is destroyed by three things: too many options, unexplained pricing, and vagueness dressed up as sophistication. If someone has to work to understand what they get, most will not do the work. They will not tell you either — they will simply leave, and you will record it as a conversion problem when it was a comprehension problem.
The test for this stage is brutally simple. Can a stranger, after thirty seconds on your page, say out loud what you sell, who it is for, and what it costs? If not, nothing further down the funnel matters.
4. Decision — they weigh it
Here the objections arrive, in a fairly predictable order: is this real, will it work for me specifically, is it worth the money, and what happens if I am wrong. Proof answers the first. Specificity answers the second. Framing answers the third. Guarantees and reversibility answer the fourth.
The common error is answering these in the wrong order — leading with a guarantee before anyone believes the thing is real, or piling on testimonials for a person who has already decided it works and is stuck on price. Objections handled out of sequence read as noise. Handled in order, they read as a conversation.
5. Purchase — they act
The most over-designed and under-examined stage. People who have decided to buy still fail to buy, and almost always for mechanical reasons: a form asking for things you do not need, an unexpected cost at the last screen, a checkout that breaks on a phone, a payment method they do not have. This stage rewards subtraction more than any other. Every field you remove is a small increase in revenue you did not have to earn twice.
6. Retention — they stay
The classic funnel stops one stage too early. Retention is not customer service and it is not a loyalty scheme; it is the stage where you find out whether the promise the other five stages made was true. It is also the only stage that makes the rest of the funnel cheaper. A customer who stays lowers what you can afford to pay for the next one, which widens the top of the funnel without you touching the top of the funnel.
Two businesses run identical acquisition: 10,000 visitors, 2% conversion, 200 customers at $100. Both make $20,000 in month one. The first has no sixth stage — customers buy once and drift. The second retains 40% into a second purchase and 25% into a third. By month six the first has made $20,000 and must buy every new customer at full price. The second has made roughly $34,000 from the same 200 people, and can now outbid the first for traffic indefinitely. Same funnel, same traffic, one extra stage.
Where funnels leak
Every funnel leaks; that is what the shape means. The useful question is whether it leaks evenly or catastrophically in one place. Even, gradual narrowing is a healthy funnel. A cliff between two stages is a specific, findable, fixable problem — and it is almost always one of four.
- The promise gap. The ad says one thing, the page says another. People arrive expecting X and are met with Y, and they leave before deciding whether Y is any good. Symptom: strong click-through, weak time-on-page, near-zero conversion.
- The comprehension gap. They want it but cannot work out what it is or what it costs. Symptom: high engagement, lots of pricing-page bounces, questions in your inbox that your page should have answered.
- The trust gap. They understand and believe it might work — for someone else. Symptom: people reach the checkout and stop, or ask for a call and never book it.
- The friction gap. They decided and the machinery got in the way. Symptom: cart abandonment concentrated on one screen, mobile conversion far below desktop.
These are diagnosable in that order, and the order matters. A trust problem cannot be fixed by simplifying checkout. A friction problem cannot be fixed with more testimonials. The most common wasted quarter in marketing is a team optimising the wrong gap very thoroughly.
Reading the numbers
One number tells you almost nothing. “We convert at 1.8%” is a fact with no action attached to it. The same funnel described stage by stage — 100% to 34% to 22% to 9% to 1.8% — tells you exactly where to spend Monday. Measure the transitions, not the total.
Three rules make the numbers honest. First, measure the step, not the outcome — you want the drop between stage three and stage four, not a monthly revenue figure that mixes everything together. Second, change one thing at a time. Repairing a funnel in three places simultaneously tells you nothing about which repair worked, and you will carry all three forward forever without knowing which one earns its keep. Third, give it enough volume to mean something. A 40% swing on 25 visitors is weather, not signal, and rebuilding a page because of it is how teams spend months going sideways.
There is a fourth rule that is less about statistics and more about honesty: write down what you expect before you look. A number you predicted and missed teaches you something about your model of the customer. A number you interpret afterwards will always seem to confirm whatever you already believed.
The two ends everyone forgets
If you take one structural idea from this piece, take this: the classic funnel is wrong at both ends, and both errors cost money in the same way.
At the top, it starts too late. The model begins at attention, as though noticing you were the same as being interested in you. It is not. Between the two sits connection — the moment a person decides this is addressed to them. Funnels that skip it are efficient at moving uninterested people toward a decision they were never going to make, which is why they can post excellent top-of-funnel numbers and terrible revenue.
At the bottom, it stops too early. Ending at purchase means the most expensive thing you own — a customer who has already paid, already trusts you, and already knows how the thing works — falls out of the model entirely. Everything you spent to acquire them is treated as consumed rather than as an asset that could pay again. Businesses in this position are not growing; they are running, and the moment traffic gets more expensive they discover there was never anything underneath.
Attention on its own converts nothing, and a funnel that ends at checkout throws away the customer it just paid to win.
Building one in order
Most funnels are built in the wrong order — traffic first, offer somewhere in the middle, measurement last if at all. Reverse it. The order below is deliberately unglamorous, and each step exists to stop you wasting the next one.
Only after those four does traffic make sense, and even then in small, deliberate amounts. The first hundred visitors exist to tell you where the funnel breaks, not to make money. Teams that skip straight to spend are buying an expensive education they could have had for a fraction of the price.
Then repeat, narrowly. Find the widest gap, form one hypothesis about why it exists, change one thing, and measure the same transition again. This is slower than a redesign and dramatically more effective, because at the end of it you know something durable about your customer rather than owning a new page you cannot explain the performance of.
How funnels break
Three failure modes account for most dead funnels. None are technical. All are discipline problems that look like strategy problems.
Optimising the wrong stage. Teams work on what they enjoy or on what is easiest to change — usually the page — regardless of where the numbers say the leak is. A funnel losing 70% of people between the ad and the page cannot be saved by improving the checkout. This is the single most common way to spend a quarter and end it exactly where you started.
Confusing motion with progress. More traffic, more channels, more variants, more tests. It feels like work and it fills a report. But volume on top of a broken middle just increases the number of people who experience your worst step. Fix the step, then increase the volume — that order, always.
Rebuilding instead of repairing. When a funnel underperforms, the instinct is to start over. It feels decisive, and it destroys the only asset you had: the data that told you where the problem was. The new funnel usually breaks in exactly the same place, because the underlying misunderstanding of the customer travelled with you. Teams that rebuild every launch are not iterating; they are looping.
A team spent six weeks redesigning their checkout after conversion fell. It did not recover. When they finally measured each stage separately, the checkout was converting at 68% — the highest in the funnel. The collapse was between the ad and the landing page, where a headline change three months earlier had quietly broken the promise the ad was making. The fix took forty minutes. The diagnosis was the whole job; the repair was trivial.
When a funnel becomes a system
A funnel that works has a ceiling, and the ceiling is arithmetic. If every customer arrives through paid attention and buys once, growth is a function of spend — and spend has a price that only ever rises. You can improve conversion rates for a while, but you are optimising within a model that structurally cannot compound.
A system changes the arithmetic in two places. It captures the audience, so that people who were not ready to buy remain reachable without paying for their attention twice. And it keeps the customer, so that the cost of acquisition is spread across several purchases instead of one. Neither is a marketing tactic. Both are structural decisions about where the value of your work accumulates — in someone else’s ad platform, or in something you own.
This is the whole distinction, and it is smaller than it sounds: the funnel does not disappear. It becomes the acquisition layer inside something larger. Everything in this article still applies — the stages, the gaps, the measurement discipline — it simply stops being the entire plan. Most businesses do not need a better funnel. They need the thing that happens after it.
You will know you have crossed the line when your growth stops being a straight function of last month’s ad spend, and when a good month makes the following month cheaper rather than merely raising the target. Until then, you have a funnel. After that, you have a system.
Key takeaways
- A funnel is a measurement device first. Its job is to tell you which step loses people. A funnel you cannot read stage by stage is a drawing, not a tool.
- Six stages, not four. Attention, connection, clarity, decision, purchase, retention — and the two the classic model omits are where most of the money is.
- Leaks are diagnosable in order. Promise, comprehension, trust, friction. Fixing them out of order is the most common way to waste a quarter.
- Measure transitions, change one thing. A single conversion rate tells you something is wrong; stage-by-stage numbers tell you where.
- The funnel is the acquisition layer, not the plan. When growth stops tracking last month’s spend, a funnel has become a system.
Conclusion
The funnel has survived a century of marketing fashion because it is honest about something uncomfortable: most people will not buy, and pretending otherwise does not change the number. Used properly, it is not a machine for extracting customers. It is a way of seeing clearly — of turning a vague sense that “marketing is not working” into a specific sentence about a specific step that specific people are leaving at.
That is the whole discipline. Name the stages. Measure the transitions. Find the widest gap. Fix one thing. Look again. It is slower than a rebuild and it compounds, because every cycle leaves you knowing something about your customer that you did not know before — and that knowledge, unlike a landing page, does not need replacing next quarter.
And when the arithmetic stops working — when growth is welded to spend and every customer costs full price — the answer is not a better funnel. It is the two stages the classic model left out. See how Funnelology maps the whole system end to end.
Frequently asked questions
A funnel is the path a person takes from first hearing about you to buying from you, split into stages so you can see where they leave. The name comes from the shape of the numbers: more people enter at the top than come out the bottom, at every single step. Before it is a marketing asset, a funnel is a measurement device. Its first job is not to sell - it is to tell you which step is losing people.
The classic model is attention, interest, desire and action. The honest modern version is attention, connection, clarity, decision, purchase and retention. The two additions are the ones that matter: connection near the top, because attention on its own converts nothing, and retention at the bottom, because a funnel that ends at checkout throws away the customer it just paid to win.
Look at the drop between each pair of stages rather than the number at the end. A single conversion rate tells you that something is wrong; stage-by-stage numbers tell you where. A working funnel has no single step where the majority of people vanish without explanation, and it has a bottom stage where customers stay long enough to be worth what you paid for them.
Nine times out of ten the promise breaks somewhere between the click and the page. Traffic proves the top works; conversion tests whether the middle keeps its word. Check that the page repeats the exact promise that earned the click, that there is one obvious next step rather than four competing ones, and that the real objection is answered before you ask for money.
The first working version should take days, not months. A funnel with one offer, one page, one follow-up sequence and honest measurement beats an elaborate one that took a quarter to build and cannot tell you where it leaks. Build the smallest complete path first, put real traffic through it, and let the numbers tell you which stage deserves the next month of work.
A funnel ends at the sale. A growth system treats the sale as a midpoint. The funnel does not disappear - it becomes the acquisition layer inside something larger that captures the audience, holds the relationship and makes the second purchase cheaper than the first. Most businesses do not need a better funnel. They need the thing that happens after it.