TOFU, MOFU and BOFU split an audience into three states of awareness: discovering a problem, comparing approaches, and choosing a supplier. The vocabulary is useful. The image it comes with, a funnel people descend through in order, describes almost nobody.

That matters because the image drives decisions. Teams build content in stage order, judge each stage with the wrong metrics, and expect a queue that never forms.

This page covers what the three stages usefully distinguish, where the model came from, what the published evidence says about its shape, and how to keep the useful half.

What the three stages actually distinguish

Strip away the geometry and what remains is a distinction about awareness, which is genuinely useful.

Top of funnel. People who have a problem and have not named it, or have named it and do not know solutions like yours exist. They are not evaluating anything, and content aimed at them competes with everything else in their day.

Middle of funnel. People who have named the problem and are working out what class of solution fits, and which suppliers are credible. This is where comparison happens, and where most B2B content is either absent or thinly disguised sales material.

Bottom of funnel. People who have chosen an approach and are reducing the risk of choosing you specifically. Their questions are about implementation, references, contracts and what happens when it goes wrong.

Why the distinction earns its place. It stops you writing one message for three audiences. Someone who does not know the category exists and someone comparing your contract terms need different things, and conflating them produces content that serves neither.

Where it starts misleading. The moment you read it as a sequence with a direction and a clock. Nothing in the evidence supports that reading, and quite a lot contradicts it.

Provenance of the marketing funnel model and its disputed attributionDiagram tracing the provenance of the marketing funnel model. The funnel descends from the AIDA sequence, standing for attention, interest, desire and action, conventionally dated to eighteen ninety-eight and attributed to E. St. Elmo Lewis, an attribution that appears in almost every textbook. A review of the primary sources published in the proceedings of a marketing history conference concludes that the dominant hypothesis crediting Lewis with inventing and formulating AIDA is unsupported or based on weak evidence, and that there is no evidence indicating Lewis invented the prototype in or around eighteen ninety-eight. The same research credits Arthur Frederick Sheldon instead, stating that Sheldon recognised the importance of the fourth word and the necessity of theorising AIDA much earlier than Lewis or any other researcher, and that Sheldon rather than Lewis can be regarded as the originator, with Frank H. Dukesmith also contributing decisively to the formulation. The foundational model of the category is therefore over a century old, was designed to train individual salespeople in face-to-face selling, and is credited to the wrong person in most of the literature that invokes it. It was never designed to describe a market, since it described the mental states of one prospect in front of one salesperson, and turning it into a population model with volumes descending between stages was a later act of translation that nobody validated.A sales script, promoted to a market modelWhat it was: AIDA, conventionally dated 1898Attention, interest, desire, action. The mental states of one prospect facing one salesperson.Who it is credited to, and who the research credits”No evidence indicating that Lewis invented the prototype.” Sheldon and Dukesmith, not Lewis.What was done to it laterTurned into a population model with volumes descending between stages. Nobody validated that step.None of this makes the model useless.All of it should lower your confidence in treating it as a law.
A sales training device for one salesperson in front of one prospect, later translated into a population model that nobody validated. Source : Proceedings of CHARM (2026)

Where the model actually comes from

The funnel is not a finding. It is a nineteenth-century sales training device that acquired a shape, and the history is more interesting than the framework.

The ancestor. The funnel descends from AIDA: attention, interest, desire, action. The sequence is conventionally dated to 1898 and attributed to E. St. Elmo Lewis, and that attribution appears in almost every textbook.

What the historical research actually found. A review of the primary sources, published in the Proceedings of the Conference on Historical Analysis and Research in Marketing, concludes that “the dominant hypothesis that E. St. Elmo Lewis invented and formulated AIDA is unsupported or based on weak evidence” and that “there is no evidence indicating that Lewis invented the prototype of AIDA in or around 1898.”

Who the same research credits instead. “Arthur Frederick Sheldon recognized the importance of the fourth word and the necessity of AIDA’s theorization much earlier than Lewis or any other researchers. Sheldon, not Lewis, can be regarded as the originator of AIDA.”

Why that is worth thirty seconds of your attention. The foundational model of the entire category is over a century old, was designed to train individual salespeople in face-to-face selling, and is credited to the wrong person in most of the literature that invokes it. None of that makes it useless. All of it should lower your confidence in treating it as a law.

What it was never designed to do. Describe a market. AIDA described the mental states of one prospect in front of one salesperson. Turning it into a population model, with volumes descending between stages, was a later act of translation that nobody validated.

Why the B2B buying journey is not a funnel

Two large pieces of published work say the same thing from different directions, and neither comes from a critic of the model.

What Google found when it looked. Its study of purchase behaviour describes the middle of the journey as a loop between two modes: “Exploration is an expansive activity, while evaluation is inherently reductive.” Buyers add options, then narrow them, then add again.

How Google characterises that loop. “This seemingly infinite construct is the defining characteristic of the messy middle.” Not a stage. Not a descent. A construct with no built-in exit, which the buyer leaves when they feel ready rather than when a stage completes.

What Gartner found about who is doing it. A survey of 632 B2B buyers, fielded August to September 2024, describes buying groups “ranging from five to 16 people across as many as four functions.”

Put those two together and the funnel breaks twice. Once because there is no single path, and once because there is no single traveller. Five to sixteen people, each running their own exploration and evaluation loop, on different timelines, arriving at different levels of conviction.

The practical consequence. Your “bottom of funnel” content is being read by someone who joined the process last week and is at the top of theirs. Your “top of funnel” content is being read by a champion who is trying to sell you internally and needs ammunition. Stage-gating your content by buyer stage assumes a queue that does not exist.

What this does not license. It does not mean the distinction between awareness states is wrong. Exploration and evaluation are real and different. It means the arrow between them is not.

The exploration and evaluation loop and the size of a business buying groupDiagram presenting two published findings that together contradict the funnel shape. The first comes from a search platform’s study of purchase behaviour, which describes the middle of the buying journey as a loop between two mental modes, exploration being an expansive activity in which buyers add brands, products and category information to their consideration sets, and evaluation being inherently reductive in which they narrow those options down. The study characterises that loop as a seemingly infinite construct which is the defining characteristic of what it names the messy middle, meaning it is not a stage and not a descent but a construct with no built-in exit, which the buyer leaves when they feel ready rather than when a stage completes. The second comes from a survey of six hundred and thirty-two business buyers fielded in August and September two thousand and twenty-four, which describes buying groups ranging from five to sixteen people across as many as four functions. Putting the two together breaks the funnel twice, once because there is no single path and once because there is no single traveller, since five to sixteen people are each running their own exploration and evaluation loop on different timelines and arriving at different levels of conviction. The practical consequence is that bottom of funnel content is read by someone who joined the process last week and is at the top of their own journey, while top of funnel content is read by an internal champion who needs ammunition.No single path, and no single travellerThe loopExploration: expansive. Options are added.Evaluation: reductive. Options are cut.Then again. In no fixed order.”A seemingly infinite construct.”The buying group5 to 16people, across up to four functionsSurvey of 632 B2B buyers,fielded August to September 2024.Each of those people is running their own loop, on their own timeline.Your bottom-of-funnel content is read by someone who joined last week and is at the top of theirs.Your top-of-funnel content is read by a champion selling you internally, who needs ammunition.
The funnel breaks twice: there is no single path, and there is no single traveller. Source : Google and Gartner (2026)

The 95:5 proposition, and what it is actually worth

You will meet this alongside the funnel, usually as the argument for spending at the top. It is directionally sound and it is not what people think it is.

The claim. At any given moment, roughly 95% of business buyers in a category are not in the market, and only 5% are.

Where it comes from. John Dawes at the Ehrenberg-Bass Institute, popularised by LinkedIn’s B2B Institute in 2021.

What its author says about it, on his own site. “The 95% figure is not meant to be a precise rule. We’re using it as a heuristic to get the idea across.”

How the number was obtained. Arithmetic, not survey. If firms change a major service provider roughly every five years, about 20% are in market across a year, and roughly 5% in a given quarter. Nobody measured the ratio itself.

Why it still earns its place in this discussion. The direction is almost certainly right, and it explains something the funnel obscures: most of your addressable market is not moving through any stage at all right now. They are not at the top of a funnel. They are outside it.

What it does not license. Quoting 95% as a measured fact, or building a budget model on the precision of the split. Use it as an order of magnitude and say so, which is what its author does.

The connection to everything above. A funnel implies a population in transit. The most defensible thing we know about B2B demand is that most of it is stationary. Those two pictures are not compatible, and the second is the one with a mechanism behind it.

What content each stage actually needs

The split below is the conventional one, and it is worth using. It is also worth knowing where it comes from.

StageWhat the content is forTypical formats
TopBeing known at allArticles, social posts, awareness video
MiddleBeing compared and kept on the listWebinars, case studies, guides, comparisons
BottomReducing risk and enabling a decisionReferences, demos, calculators, proposals

Who publishes this grid, and why that matters. The most cited version comes from a company selling CRM and marketing automation software whose business model is built on the funnel: attract, nurture, then convert leads inside its tool. That does not make the definitions wrong, and they are genuinely well written. It explains the conviction.

A detail that gives the game away. The same source acknowledges that top and middle content overlap, because prospects do not have a clear view of their own problem. Even the vendor selling the model concedes the stages leak.

The mistake this grid produces most often. Building in stage order, top first, on the theory that you need an audience before you need anything else. In B2B, the reverse is usually right: the bottom-of-funnel pages, the ones answering “what does it cost”, “how does it work”, “who else uses it”, are the pages your existing demand is already looking for and failing to find.

A better sequencing question. Not “which stage comes first” but “which questions is somebody already asking us that we have never answered in writing”. That list is almost always bottom-heavy, and it is the cheapest content you will ever produce.

Which metrics belong to each stage

A warning first, because it is more useful than the list. There is no normative table establishing that a given metric belongs to a given stage. What follows is a sensible convention consistent with the logic of the journey, not a sourced finding, and we present it as such.

Top: reach. Impressions, reach, traffic, new visitors, share of voice, brand recall. You are measuring presence, not sales.

Middle: engagement. Leads, qualified leads, click-through rate, cost per lead, webinar registrations. You are measuring interest turning into contact.

Bottom: conversion. Conversion rate, qualified meetings, win rate, pipeline value, acquisition cost. You are measuring decision and its cost.

The rule that matters more than the list. Never judge one stage with another stage’s metrics. An awareness campaign assessed on acquisition cost will always look disappointing, because converting is not its job. A bottom-of-funnel budget assessed on reach will miss the only thing it was for.

Where this goes wrong in practice. A quarterly review puts every channel in one table with cost per lead as the comparison column. The awareness work loses, gets cut, and the pipeline thins two quarters later with no visible cause. This is the single most common way a B2B marketing budget destroys itself, and it takes one badly built table.

Choosing what to run the budget on. Once all these measures are on the table, deciding which one governs the budget is its own arbitration, and we settle it in ROAS vs MER vs CAC vs LTV.

Metrics conventionally attached to each funnel stage and the rule governing their useTable setting out the metrics conventionally attached to each stage of the marketing funnel, with the warning that no normative table establishes which metric belongs to which stage, the split below being a sensible convention consistent with the logic of the journey rather than a sourced finding. At the top of the funnel the measure is reach, comprising impressions, reach, traffic, new visitors, share of voice and brand recall, which measures presence rather than sales. In the middle the measure is engagement, comprising leads, qualified leads, click-through rate, cost per lead and webinar registrations, which measures interest turning into contact. At the bottom the measure is conversion, comprising conversion rate, qualified meetings, win rate, pipeline value and acquisition cost, which measures decision and its cost. The rule that matters more than the list is never to judge one stage with another stage’s metrics, because an awareness campaign assessed on acquisition cost will always look disappointing given that converting is not its job, while a bottom of funnel budget assessed on reach will miss the only thing it was for. Where this goes wrong in practice is a quarterly review that places every channel in one table with cost per lead as the comparison column, whereupon the awareness work loses, is cut, and the pipeline thins two quarters later with no visible cause.One table, one column, and the awareness work losesTop: reachImpressions, traffic, share of voice, brand recallMeasures presence. Judged on cost per lead, it always loses.Middle: engagementLeads, qualified leads, cost per lead, registrationsMeasures interest turning into contact.Bottom: conversionConversion rate, meetings, win rate, acquisition costMeasures decision and its cost.No normative table establishes this split. It is a convention, and we present it as one.The rule that matters: never judge one stage with another stage’s metrics.
The most common way a B2B marketing budget destroys itself: one table, one comparison column, and the awareness work loses. Source : MASTRATOS (2026)

This funnel is not your website’s funnel

Two different objects share the word, and mixing them produces months of confused reporting.

The marketing funnel describes states of awareness across a market. It has no fixed duration, no fixed order, and its population is everyone who could eventually buy.

The website conversion funnel describes a sequence of steps on your own site: page, form, thank-you page. It is short, ordered, and its population is people already on your site this session.

Why conflating them breaks reporting. The second is measurable to the step. The first is not measurable at all in the same sense, because you cannot observe the awareness state of people who never arrived. Treating a marketing funnel like a website funnel produces drop-off rates between stages that nobody actually measured.

The stage-to-stage conversion rates you will be shown. Percentages moving from top to middle, middle to bottom. Ask how the denominator of the first one was counted. There is no observable population of “people at the top of the funnel”, so the number is a construction, and usually a backwards-derived one.

What to do instead. Measure the website funnel properly, step by step, because you can. Treat the marketing funnel as an allocation framework, not a measurement one. The two documents should look different and answer different questions.

Difference between the marketing funnel and the website conversion funnelDiagram distinguishing two different objects that share the word funnel and whose conflation produces months of confused reporting. The marketing funnel describes states of awareness across a market, has no fixed duration and no fixed order, and its population is everyone who could eventually buy. The website conversion funnel describes a sequence of steps on the company’s own site, such as a page, a form and a thank-you page, is short and ordered, and its population is people already on the site during that session. Conflating them breaks reporting because the second is measurable step by step whereas the first is not measurable in the same sense at all, since the awareness state of people who never arrived cannot be observed, so treating a marketing funnel like a website funnel produces drop-off rates between stages that nobody actually measured. When presented with stage-to-stage conversion percentages moving from top to middle and middle to bottom, the question to ask is how the denominator of the first one was counted, since there is no observable population of people at the top of the funnel and the number is therefore a construction, usually derived backwards. The recommendation is to measure the website funnel properly step by step because that is possible, and to treat the marketing funnel as an allocation framework rather than a measurement one, the two documents looking different and answering different questions.Two objects, one wordThe marketing funnelStates of awareness across a marketNo fixed duration, no fixed orderPopulation: everyone who could buyNot measurable step by step.You cannot observe people who never arrived.The website funnelPage, form, thank-you pageShort and orderedPopulation: people on your site nowMeasurable to the step.Measure this one properly, because you can.Shown a stage-to-stage conversion rate? Ask how the first denominator was counted.There is no observable population at “the top of the funnel”. The number is a construction.
One is measurable step by step because you can observe every step. The other is not measurable in that sense at all. Source : MASTRATOS (2026)
What to keep and what to drop from the funnel modelTable setting out what to keep and what to drop from the marketing funnel model in order to use it without being trapped by it. What to keep includes using it for allocation, since deciding what share of budget and content serves people who do not know the company against people choosing between it and two competitors is a real decision that the three stages frame well, and using it for message discipline, since writing for one awareness state at a time produces sharper content than writing for everyone. What to drop includes the sequence, meaning not building top of funnel first but instead building the pages that answer questions people already ask and then working outward, because demand that already exists is cheaper to serve than demand that must be created. It also includes stage gating, meaning not withholding pricing, references or technical detail until a prospect is judged far enough down, since buying groups of five to sixteen people include somebody who needs that detail today to argue the case internally. And it includes the descent metaphor in reporting, meaning ceasing to present stage-to-stage conversion rates that cannot be computed and presenting absolute counts per stage instead, letting the reader see the shape without a percentage that implies a measurement nobody made. The single line to retain is that the stages describe states rather than a queue.Keep the distinction, drop the geometryKeepUsing it to allocate budget and contentWriting for one awareness state at a timeNever judging a stage on another’s metricsThese are real decisions, and the threestages frame them well.DropBuilding top-firstGating pricing and detail by “stage”Stage-to-stage conversion percentagesEach of these assumes a queue thatdoes not form.The stages describe states, not a queue. Everything useful survives that correction.
Everything useful about the model survives the correction. Most of what misleads does not. Source : MASTRATOS (2026)

Using the model without being trapped by it

Keep the distinction, drop the geometry. That is the whole recommendation, and here is what it looks like in practice.

Keep it for allocation. Deciding what share of budget and content serves people who do not know you against people choosing between you and two competitors is a real decision, and the three stages frame it well.

Keep it for message discipline. Writing for one awareness state at a time produces sharper content than writing for everyone, which is what most B2B pages currently do.

Drop the sequence. Do not build top-first. Build the pages answering questions people already ask you, then work outward. The demand that exists is cheaper to serve than the demand you have to create.

Drop the stage gating. Do not withhold pricing, references or technical detail until someone is “far enough down”. Buying groups of five to sixteen people include someone who needs that detail today to argue your case internally.

Drop the descent metaphor in reporting. Stop presenting stage-to-stage conversion rates you cannot compute. Present absolute counts per stage and let the reader see the shape without a percentage that implies a measurement nobody made.

The one line to keep from all of this. The stages describe states, not a queue. Everything useful about the model survives that correction, and most of what misleads does not.

Where to go next

You are choosing which metric governs the budget. ROAS vs MER vs CAC vs LTV.

Your website funnel is the one you want to fix. What makes a B2B landing page convert, and conversion rate.

You are splitting brand and performance budget. Brand vs performance in B2B.

Your bottom-of-funnel number is the problem. Cost per lead, and CAC at low volume.

You want the buying side of the chain. Media buying explained.

In short

  • The three stages describe states of awareness, and that distinction is genuinely useful for allocation and for message discipline.
  • The funnel descends from AIDA, conventionally dated 1898, and historical review of the primary sources finds the usual attribution to Lewis “unsupported or based on weak evidence”.
  • It was a script for one salesperson facing one prospect. Turning it into a population model with volumes descending between stages was a later translation nobody validated.
  • Google’s own research describes a loop, not a descent: exploration is expansive, evaluation is reductive, and the loop is “a seemingly infinite construct”.
  • Buying groups run from five to 16 people across up to four functions, so there is no single traveller either.
  • No normative table assigns metrics to stages. The convention is sensible; treat it as a convention.
  • Never judge one stage with another stage’s metrics. One table with cost per lead as the comparison column is how awareness budgets get cut.
  • The marketing funnel is not your website funnel. One is measurable step by step. The other is not measurable in that sense at all.

The stages are a way of allocating attention, not a description of how anyone buys. Book a diagnostic, or see how we approach B2B paid acquisition.