Every business is a funnel

I have been writing a guide to fundraising lately, and somewhere in the middle of it I noticed that I kept drawing the same picture. Fundraising, it turns out, is a funnel: forty or sixty well-matched investors at the top, a dozen meetings in the middle, two or three serious conversations near the bottom, and, if the process is run well, one signed term sheet at the spout. Then I noticed the picture was not really about fundraising at all. Sales is the same picture. Hiring is the same picture. A shop, a restaurant, a newsletter, a church plant, a freelancer's pipeline of clients: the same picture, relabelled. That is the claim of this essay, meant literally. Every business, described honestly, is a funnel, and most of what goes wrong in a business is easier to name, easier to fix, and much easier to endure once you say so.

The shape

Here is the shape, stripped of jargon. At the top stands everyone who could conceivably care about what you do: a wide, indifferent crowd of strangers. At the bottom stands a much smaller group: the people who pay you, come back, and tell their friends. Between the two run a series of narrowing stages, and each stage is a decision made by a person, not by you. A stranger becomes aware of you. Some of the aware become interested enough to look closer. Some of the interested come to trust you enough to try. Some of the triers pay. Some of the payers return, and some of the returners recommend. At every boundary, most people fall away. That is not the failure of the design; it is the design. A funnel that did not narrow would be a pipe, and there has never been a business, not one, whose every passer-by became a devoted customer.

“There is only one valid definition of business purpose: to create a customer.”

— Peter Drucker, The Practice of Management, 1954

Drucker's definition is usually quoted as a corrective to financial navel-gazing (the purpose is not profit; profit is the test), but read it structurally and it says something more useful. If the purpose of a business is to create a customer, then the funnel is the anatomy of that creation: the sequence of states a stranger passes through on the way to becoming one. A customer is not found, like a coin on the pavement. A customer is made, stage by stage, out of a stranger's attention, then their interest, then their trust, then their money, then their habit. Every business is in the business of running people through that transformation, whatever it says on the sign. The corner bakery runs the funnel with a window display and the smell of bread; the enterprise software firm runs it with whitepapers and a sales team; the startup founder raising a round runs it with warm introductions and a deck. The labels change. The shape does not.

The arithmetic

The first gift of the funnel is diagnostic. A struggling business almost never knows, precisely, what is wrong with it. It knows how the trouble feels: revenue is down, things are quiet, we need more customers. But “we need more customers” is not a diagnosis; it is a symptom reported in a tone of voice. The funnel turns the feeling into arithmetic. Your revenue is a multiplication: the number of people who hear of you, times the fraction who look, times the fraction who try, times the fraction who buy, times what they pay, times how often they come back. Write the stages down, put real numbers against them (even rough ones), and the vague dread resolves into a location. It is no longer “business is bad.” It is: plenty of people walk in, and almost nobody buys. Or: everyone who tries us loves us, and nobody has heard of us. Or: they buy once and never return. These are three different diseases. They have three different cures. Before the funnel, they present as the same ache.

“An hour lost at a bottleneck is an hour lost for the entire system.”

— Eliyahu Goldratt, The Goal, 1984

Goldratt was writing about factories, but the theorem transfers whole. In any chain of dependent stages, the system's output is set by its narrowest point, and effort spent widening any other point is effort wasted. This is the funnel's second gift, and it is a severe one: it tells you that most of the things you could do to improve your business will do nothing at all. The restaurant whose problem is that nobody walks past does not need a better menu. The app whose problem is that nobody who downloads it survives the first five minutes does not need more advertising; more advertising just pours more water over the same hole. The discipline is to find the stage where the numbers collapse, fix that stage, and leave the others alone until it is fixed. This sounds obvious. It is obvious. It is also, in my experience, almost never done, because the bottleneck stage is usually the uncomfortable one (the pitch that isn't landing, the product people quietly abandon), and the stages we polish instead are the ones we enjoy.

This is what the diagrams above are actually for. A funnel drawn on paper is an instrument panel waiting for its needles: write last month's count at every boundary and each adjacent pair of numbers becomes a rate, each rate has a rough going benchmark, and the bottleneck stops being a matter of opinion. If forty thousand people saw the app's store page and three hundred downloaded, the problem is not awareness, and no amount of keyword work will touch it; the page is failing people who already arrived. If everyone in the coffee shop reads the menu and few reach the till, the problem is not the signage; it is the menu or the prices. This is the quiet payoff of the clear approach: arguments about “the business”, which are unwinnable because the business is everything at once, become readings at a boundary, which are checkable. The meeting that used to trade adjectives now points at a number, and the number points at a stage, and the stage has an owner and a fix.

The dignity of no

The funnel's third gift is not analytic but pastoral, and for anyone actually running the thing it may matter most. A funnel prices rejection in. If forty investors yield one term sheet, then thirty-nine of the best-case outcomes are refusals, which means a single “no” carries almost no information. It is not a verdict on you, your product, or your life's worth. It is one draw from a distribution whose whole shape you already accepted when you drew the funnel. The salesman who knows his numbers close at one in twenty does not experience the nineteenth refusal as tragedy; he experiences it as progress, the necessary spending-down of the noes that stand between him and the yes. The founder who does not know her numbers experiences every pass as a little death, and twenty little deaths in a row will end a fundraise, and sometimes a company, that the arithmetic said was going fine.

The meaningful unit is never the single answer; it is the rate. One “no” means nothing. A hundred noes at a stage where the comparable rate is one in five means something exact, and what it means is not give up but this stage is broken; look here. The funnel, taken seriously, is therefore an instrument of morale as much as of measurement. It converts rejection from a judgment into a cost of goods: something budgeted, expected, and survivable. There is a reason every durable sales culture teaches its people the pipeline before it teaches them the pitch. You cannot keep a person knocking on doors with encouragement alone. You can keep them knocking with a denominator.

Leaks and narrowing

A distinction the picture makes easy: a funnel is supposed to narrow, but it is not supposed to leak, and the two are different events that call for opposite responses. Narrowing is the healthy loss of people for whom the thing was never right: the browser who was only sheltering from the rain, the investor whose fund does not touch your sector, the customer whose problem you honestly do not solve. Losing them is not merely acceptable; it is qualification working, and trying to keep them is how businesses blur themselves into meaning nothing. A leak is the other thing: losing a person who wanted to continue and was prevented. The checkout that fails on a phone. The email that goes unanswered for a week. The signup that demands eleven fields before it shows the product. The restaurant with a queue out the door and no one taking names. Narrowing is the design doing its job. A leak is the design failing someone who was trying to say yes.

Confusing the two produces the most expensive mistake in the funnel business: widening the top to compensate for a broken middle. It is always available (attention can always be bought), it always works a little (some fraction of the new water makes it past the hole), and it is almost always wrong, because you are paying full price for strangers in order to lose them at a stage you could have fixed once, for good. The sound rule runs the other way: fix the funnel from the bottom up. A repaired bottom stage multiplies the value of every person already entering the top; a widened top multiplies the waste of every leak below it. Retention before conversion, conversion before traffic. Only when the lower stages hold water is a wider top worth what it costs.

Iterate the stages

A funnel drawn once is a diagnosis. A funnel drawn every quarter is a discipline, because the bands are not scenery. Each one is a product in its own right, with its own users, its own job and its own measurable rate, which means each one can be improved by exactly the loop that improves the product at the centre: change the stage, watch its boundary, keep what converts, and go again. The store page is a product. The onboarding is a product. The menu, the till queue, the renewal email, the cancellation flow: products, every one, each capable of serving people better this quarter than it did last. Drucker drew the line under all of this in 1954, in the sentence that follows the one this essay started from:

“Because its purpose is to create a customer, the business enterprise has two—and only these two—basic functions: marketing and innovation.”

— Peter Drucker, The Practice of Management, 1954

Read against the funnel, the pairing is exact. Marketing is running the funnel; innovation is refining it; everything else in the building is overhead in service of the two. And the refining is meant to be a repeatable process, not an occasional heroic redesign: a steady cadence of small, deliberate improvements, stage by stage, that compound the way conversion rates multiply. Apple is the patient example. Over two decades it has rebuilt nearly every band by hand: the retail store is a reinvented middle (walk in off the street and handle the machine, trial without commitment, questions answered by someone not paid on commission); the unboxing is the first minute of the Stay band, engineered with the care other firms reserve for the product itself; and the ecosystem is the Advocate band industrialised, each device making the next one easier to say yes to. None of it arrived at once. It accreted, release by release, each generation of each stage slightly better at its one conversion, until the whole funnel felt inevitable.

Tesla and SpaceX are the impatient examples. Tesla deleted a band it did not own: the dealership, a leaky middle stage staffed by intermediaries with their own incentives, replaced by an order form that takes ten minutes. Then it did something stranger, and iterated a stage most businesses treat as finished: the car improves after purchase, over-the-air updates working on the Stay band while the thing sits in the drive, so that ownership itself keeps converting toward advocacy. SpaceX runs the same loop below the funnel entirely. Every launch, and especially every failure, is a turn of build–measure–learn at rocket scale, and the iteration surfaces in the funnel as a widening top: each fall in the cost of a launch changes who can afford to be a customer at all, so a market that consisted of governments came to include broadcasters, startups, and eventually a university with a shoebox satellite. That is the general lesson of all three. Iterate the stages and the service improves; iterate the machine underneath and the funnel improves everywhere at once, because better service is upstream of every rate in it.

The bottom feeds the top

There is one more thing the picture shows, and it is the difference between a business that must be pushed and one that pulls. In a mediocre business the funnel is a line: strangers in at the top, customers out at the bottom, and every stranger paid for. In a good one the funnel bends into a loop, because the people at the bottom (the ones who love the thing) become the top's cheapest and most credible source. A delighted customer is not an output. She is an input: her recommendation lands with a trust that no advertisement can buy, entering the funnel three stages deep, past awareness and interest and most of the way through scepticism. Word of mouth is not a marketing channel among others. It is the bottom of the funnel reaching up and refilling the top.

“It's better to have 100 people who love you than a million people that sort of like you.”

— Paul Graham, to the first Airbnb founders

This is why the advice sounds paradoxical and isn't. A million who sort of like you sit inert in the middle of the funnel, converting weakly, referring nobody. A hundred who love you are a hundred small engines pumping new strangers in at the top, and love, unlike advertising, compounds. The businesses we describe as not needing marketing (the restaurant you must book a month out, the tool every developer tells every other developer about) are not funnels that somehow escaped the shape. They are funnels whose bottoms feed their tops so well that the paid top became optional. The loop is the prize. Everything else in the essay is in its service.

This is why the funnels in this essay run one band past the place where most dashboards stop. Beyond Stay sits Advocate, because a customer who merely keeps paying and a customer who loves you are different creatures, and the difference is what they do with their mouths. The advocate spends their own credibility on you: brings a friend, leaves the review nobody asked for, defends you in a comment thread, forwards the newsletter. Advocacy is the one stage whose output lands in other people's funnels rather than your own ledger, which is exactly why ledgers forget it, and exactly why it matters more than anything the ledger holds. A business whose advocate band is thick eventually notices something strange: selling has gone quiet. Drucker described that end state too, and it reads like a description of the loop closing:

“The aim of marketing is to make selling superfluous. The aim of marketing is to know and understand the customer so well that the product or service fits him and sells itself.”

— Peter Drucker, Management: Tasks, Responsibilities, Practices, 1973

A product that sells itself has not escaped the funnel. It is a funnel whose advocate band has grown large enough to feed the top unassisted, so that the machinery of persuasion can idle. That is the state every stage of the essay has been pointing toward, and it is reached from the bottom, never from the top: not by shouting to more strangers but by serving the people already deepest in the funnel so well that they take over the shouting.

Capture, for better and worse

Everything so far has assumed that people stay at the bottom of a funnel because they want to. There is a second way to keep them: make leaving expensive. Call it capture. From inside the spreadsheet the two are indistinguishable (retention is high either way), but they are opposites in kind, because one is wanting measured and the other is wanting made irrelevant. And whole sectors run on captured funnels. Education captures by sunk years: nobody re-chooses their university in the third year of a degree, whatever the teaching turns out to be, because exit means repaying the toll in time. Health captures by urgency and opacity: much of that funnel is entered on a stretcher, and the patient rarely chooses, compares, or even sees a price. Government is the limit case, a monopoly provider whose churn mechanism is emigration, which is why its counters so seldom feel like a bid for your custom. And the large technology platforms have made capture a science: two phone operating systems, two app stores standing at the top of every app's funnel and taking a toll on its bottom, and ecosystems whose exit price (your photo library, your purchases, your group chats) is set carefully at just above what leaving would be worth.

The trouble with capture is not only moral. It destroys the funnel's information. The whole argument of this essay is that the funnel is an instrument: each stage measures a kind of wanting, which is what makes the numbers worth reading. A captured bottom measures nothing. Retention stays green while the love drains out, so the organisation reads loyalty where there is only lock-in, and because churn can no longer discipline quality, quality decays on schedule: the platform, safe from exit, begins quietly redirecting value from the people who cannot leave to itself. (Cory Doctorow coined an indelicate word for this decay; the mechanism is exactly a captured funnel being farmed.) Albert Hirschman mapped the territory half a century ago: a decaying organisation is corrected by exit or by voice, and the two are linked, because complaint is only listened to when the complainer could walk.

“As a rule, then, loyalty holds exit at bay and activates voice.”

— Albert O. Hirschman, Exit, Voice, and Loyalty, 1970

Hirschman's word is the right one for the benevolent version, because there is a benevolent version: capture earned rather than enforced. Loyalty is what a captured funnel looks like when the door is open. I see it at small scale with my own apps: someone who comes to love one of them does not meet the next one as a stranger. The second app enters their funnel three stages deep, past awareness, past scepticism, most of the way through trust, exactly as a friend's recommendation would, except that here the trusted friend is the previous product. That is what a brand is: stages pre-crossed, a privileged starting position at the top of the next funnel, earned by the last thing keeping its promise. It is capture of a kind (a competitor now has to beat not just my app but my record), and it is entirely benign, because it is held by deserving and renewed at every release.

The test that separates the two captures is the open door. Would they stay if leaving cost nothing? Earned capture passes, and tends to advertise the door (easy cancellation, data export, no contract), because confidence in the product is part of the pitch. Enforced capture needs the door locked, and you can read the lock straight off the interface: the cancellation buried four menus deep is a confession that the product no longer expects to be chosen. Earned capture also polices itself in a way the enforced kind never has to: it lasts exactly as long as the deserving, and betrayal runs the loop in reverse, converting accumulated trust into accumulated resentment with all the efficiency of word of mouth. A small business should notice that the benevolent kind is the only capture on offer at its scale. You cannot bar the exits. All you can do is make the next yes easier than the first one was, which is, on reflection, the only capture worth having.

What the funnel is made of

Now the caution, because a picture this useful will, if unwatched, quietly become a philosophy. The funnel describes people in aggregate, and it is easy to slide from measuring the aggregate to treating the people as liquid: units of throughput to be piped, squeezed, and optimised. But nothing flows in a funnel. At every boundary there is a person, weighing whether you have yet earned the next unit of their trust, and “conversion” is just the ledger's name for the moment they decide you have. This means the only durable way to improve a stage is to deserve the yes at that stage: to be genuinely worth noticing, worth trying, worth paying, worth returning to. The other way (the dark patterns, the countdown timers that reset at midnight, the cancellation buried four menus deep) does move the numbers, briefly, at the stage where it is applied. Then it poisons the one stage that cannot be bought. Nobody recommends the company that tricked them. The trick converts the middle and kills the loop, and the loop was the prize.

So the funnel is a description, not a strategy. It tells you where the work is; it cannot do the work, and it cannot tell you what the work is for. The startup canon compresses the strategy into four words (make something people want), and the funnel is simply that commandment differentiated: at each stage, the version of wanting that stage tests. Awareness tests whether you are worth a glance. Trial tests whether the promise survives contact. Retention tests whether the want was real or manufactured. Referral tests whether the want is strong enough that someone will spend their own credibility on it. A business that is failing at a stage is being told, precisely, which part of the wanting it has not yet earned. That is not the funnel judging you. That is the funnel handing you, with unusual courtesy, the exact address of the truth.

If you are starting one

Everything above assumes a funnel already exists. If you are at the very beginning, with nothing but an idea and the pull of the thing itself, the picture reorders your priorities in a specific way. Start from the passion and stay true to it: I have argued elsewhere that only a real vocation survives the years of tending a business demands, and a funnel built around work you do not love is a machine you will abandon at the first hard winter. But do not let the passion postpone contact. The most valuable thing you can do early is connect the work to the top of a funnel, because the top is where the market answers its first and largest question: is there an audience at all? Passion tells you the work is worth doing. Only the funnel can tell you whether anyone else agrees, and it tells you fast.

Be precise about when a funnel goes live, because it is earlier than it feels. Not launch day, not the press release, not the finished product. A funnel is live the moment one human being finds the thing and decides whether to try it. That single arrival is already data: someone saw the promise and crossed, or saw it and walked, and either way you have stopped guessing and started reading. From the first visitor onward, every boundary is reporting. (This immediacy, incidentally, is why the advertising networks grew into some of the most powerful businesses on earth: they own the tap at the top of everyone else's funnel and rent it by the click, and a tap you can switch on this afternoon, with verdicts flowing back by evening, is worth almost any toll.)

Read the top honestly, because its two earliest messages point in opposite directions and both are easy to misread. A long silence at the top is the market speaking. If the funnel is genuinely connected (the listing live, the keywords honest, the sign on the street) and week after week almost nobody comes to stand at the entrance, do not count on the crowd turning up later. Prolonged silence usually means the demand does not exist, and no amount of polishing the lower stages will conjure strangers a market never contained. The opposite reading is just as important, because it rescues founders who are about to quit for the wrong reason. A rough, unfinished product with even a handful of people trickling into the funnel, trying it, and coming back is not failure at small scale; it is one of the best signals a young business can receive: pull, however faint, where there could have been nothing. Smallness at the start is normal, and fixable. Silence is the warning. A trickle is a reason to keep going.

What the early verdicts usually say, when people are arriving but not crossing, is not give up. It is the product has not finished persuading, and often simply there needs to be more product. At a scale of dozens of users rather than thousands, the way to find out which stage is failing is not analytics, which needs volume you do not have; it is conversation.

“The most common unscalable thing founders have to do at the start is to recruit users manually.”

— Paul Graham, Do Things That Don’t Scale, 2013

Graham's advice is funnel work in disguise. Recruiting users manually is walking each one through the funnel by hand: you watch, in person, exactly where their attention snags, where their trust hesitates, where their wallet stays shut, and each walked-through customer teaches you what a stage must eventually do on its own. Doing things that don't scale is how you learn what to build into the stages so that one day they do.

The connectors themselves are industry-specific and mostly boring, which is a virtue, because boring means proven. An app plugs its top into the store: ASO and a handful of honest keywords, the product page treated as the conversion surface it is, store search ads for pennies of demand data, and a landing page that can take a waitlist before the build is finished. Software sold to businesses plugs into search and content, a launch on the forums its buyers actually read, and unglamorous cold outreach. A café or a shop plugs into the Google Maps listing, the signage, the A-board and the local feeds. A maker plugs into a marketplace that already has traffic (Etsy, Amazon) before spending a year on a site of their own; a newsletter into its platform's recommendations and swaps with adjacent writers. The principle under every case is the same: rent or borrow a top that already has people flowing past, connect it honestly to what you have made, and then work the arrivals stage by stage toward the first sale. And mind the manner of the crossings, because the manner is the brand. Every boundary crossed on earned trust compounds toward the advocate band; every one crossed by a trick is borrowed against it. The first sale matters much less than how it was made, because you are not really building sales. You are building the machine that makes them.

Funnel flow

The state all this effort is aiming at deserves a name. A funnel in which every stage does its job stops behaving like a filter and starts behaving like a passage: strangers enter at one end and advocates emerge at the other, continuously, with the loop refilling the top as fast as the bottom converts. Call it funnel flow. It is close to what the startup canon calls product–market fit, seen from the side and stretched across the whole machine: fit is the wanting, flow is the wanting successfully conducted from one end of the funnel to the other. The funnel still narrows, as it must, but what is lost at each boundary is the narrowing you designed, not the leaks you didn't, and the movement never stalls: attention becomes trial becomes habit becomes advocacy in one connected motion, and the business stops lurching between famine and campaign and simply runs.

Reaching flow can take incredible effort, years of the stage-by-stage iteration described above, and that is exactly why the rewards are disproportionate. The arithmetic multiplies: a funnel healthy at every boundary compounds while a competitor's leaks, so the same spend on the same top yields several times the customers. And flow is defensible in a way features never are, because it is not one copyable idea but dozens of small, hard-won fixes accumulated across every stage; a rival can clone your product in a quarter and still take years to clone your funnel. This is also the honest explanation of businesses that look effortlessly successful: the effort was real, and front-loaded, and is now invisible inside the flow.

But flow is a state, not a possession, and it is held only by the same adaptation and innovation that built it. Stand still and the funnel clogs. Marketing gone wrong clogs the top: a campaign that draws a crowd who were never going to convert, a message that curdles the brand it was meant to build, a rented channel whose price quietly triples. A decaying product, or the codebase underneath it, clogs from the inside: releases slow, small bugs erode the Stay band one irritation at a time, and the funnel starves at the far end first, where the damage takes longest to show up in the numbers. Prices go stale; platforms change their algorithms and a top you rented shuts overnight. There is no general prescription for any of this, because how a clog is cleared is case by case: the factors are too many and too particular for a rule. What is general is the guard duty. Watch the rates, treat a moving one as a smoke alarm rather than a curiosity, and trust the diagram: the funnel will always tell you where the trouble is, even when it cannot tell you the fix.

Name your stages

The practical discipline, then, fits in a paragraph. Draw your funnel. Name the stages a stranger actually passes through on the way to becoming your customer, in your business, not the generic ones from a textbook. Put numbers on the boundaries, even embarrassing, approximate numbers, because an approximate number beats a precise mood. Find the narrowest point and work there, resisting the polish you would rather apply elsewhere. Distinguish the narrowing from the leaks: bless the first, fix the second, and do not buy a wider top until the bottom holds water. Let no single “no” carry meaning it has not statistically earned. And keep the bottom sacred, because the people who love you are the top of next year's funnel, and everything that betrays them for a quarterly number is eating the seed corn. Do all of it, patiently and for long enough, and the funnel stops filtering and starts flowing.

I said every business is a funnel, and I have not found the exception. The job hunt is a funnel; so is the search for a co-founder, and the directory of investors is a funnel's top waiting to be worked. But the point of the claim is not the universality; it is the relief. The funnel takes the two heaviest burdens of running anything (not knowing what is wrong, and taking every refusal personally) and replaces them with a diagram: here are your stages, here is the one that is broken, here is how many noes a yes costs at the going rate. What remains after the diagram is only the real work, which no diagram does: making the thing at the bottom of the funnel so good that the people who reach it turn around and hold the top open for everyone behind them.