In Funnel Design the Number of Stages Varies by Trade, and Only the Order Stays Fixed

Nobody can answer the question “how many stages should I have?” without knowing the trade it refers to. Say four, and the four-stage template gets borrowed. Say it depends on your business, and no method of deciding has been handed over. Neither answer leaves anything in the hands of the person who asked.

The reason no answer arrives is that the question treats the number of stages as a design variable on this side. A design variable can be chosen. About things that can be chosen, the question “how many is best?” makes sense.

In fact the number of stages is not something chosen here. Where does the quantity of involvement a buyer is paying towards you change? The count of those changing points is the count of stages, and it is a property sitting on the side of the trade being conducted. In a transaction that cannot proceed without somebody else’s approval, the quality of involvement differs before and after that approval. In a transaction settled by one person, that changing point does not exist.

Funnel design is the task of observing those changing points and giving them names. It is not the task of arranging what is for sale in order of price. Why arranging by price spread as the standard has its reason in the history of the diagram. What the funnel shape brought in with it is treated in marketing design as a whole.

This distinction bites because it changes the input to the design. If the task is to arrange products, the input becomes a list of what you are able to make. If the task is to observe the shifts in involvement, the input becomes what is happening on the other person’s side. The same diagram may be drawn either way, and what is being looked at while drawing it differs.

That difference does not show up in the appearance of the diagram. Whichever procedure produced it, the result is that shape — wide at the top, narrow at the bottom. Because the shape is identical, a diagram drawn from the wrong input looks exactly like one drawn correctly.

Three things get answers here. What is actually arranged in the sequence. Where the number of stages is settled. And how to confirm whether a stage exists at all.

Some things do not get answers. How many stages your own business has will not come out here. That value emerges from observation, and the observation exists only on your side. Nor is the line of how much to hand over at each stage treated. That line is settled not by involvement but by the nature of what is being handed over.

Borrow the number of stages from a template and measures get built for stages that do not exist. They miss, and the miss is read as a fault in the measure. Rebuilding does not help. That the reason for the failure sits in the stage is invisible from the side of the measure. And with every rebuild, that stage stays in the design as one that still has room for improvement.

The Number of Stages Usually Matches the Number of Things on Hand

The circulating procedure is accurate as observation and as practical advice. It also carries a single shared premise.

A marketing funnel is the process running from a prospect’s first awareness of a product to the moment of purchase, divided into stages and given a structure. The most basic form has four: awareness, then interest, then comparison and consideration, then purchase. Numbers fall from top to bottom, which is why it is drawn as an inverted triangle.

The design procedure presented alongside it runs roughly as follows. First, analyse the target for your own business. Second, customise the stages to match that purchasing process. Third, assign to each stage the measures suited to it. Fourth, measure the figures stage by stage and improve them.

On customising the stages, there are specific observations too. In business-to-business transactions, the sequence problem recognition, then information gathering, then comparison, then internal approval, then contract sits closer to the reality. A reason is supplied with it: the purchasing decision does not conclude with one person, so a stage for approval becomes necessary.

Forms that do not end at purchase are introduced as well. A sequence that raises satisfaction after the purchase and connects into repeat business and referral, handled as a separate funnel.

These are accurate. Thinking in divided stages is itself what makes the state of a business visible. Without the division, everything gets called by the single phrase not selling, and nothing settles about where to touch.

Set alongside each other, the four procedures share one premise. All of them treat a stage as a place to assign a measure to. Advertising for the awareness stage, content for the interest stage, case studies for the comparison stage, an incentive for the purchase stage. The stage comes first and the measure enters it.

That order is not invalid. In a business that already has traffic and can count the numbers at each stage, this order works fastest. Add a measure to a stage that is producing figures, and the count passing through that stage moves.

It breaks when the order is started from before any traffic exists. With nothing to measure, the number and names of stages get borrowed from the general template. Assign measures to borrowed stages, and what results is a state where there are as many stages as there are measures. Somebody holding four measures has a four-stage funnel; somebody holding seven has a seven-stage one.

Line up a few of the diagrams presented as sales funnel design and this correspondence can be confirmed by eye. The number of stages is settled not by something on the other person’s side but by what is on hand here. That is where the design is weak as a starting point.

And the weakness does not appear on the diagram. Four stages decided from what was on hand and four stages taken from observation produce exactly the same picture. The origin of the input leaves no trace on the shape of the output.

Build Stages Out of Products and the Reason for Rising Becomes Has Not Bought Yet

In one-person businesses, filling the contents of each stage with products is especially widespread. A free document, a low-priced trial, the main product, a continuing contract. Arranged in order of price, growing more expensive further down.

The advantage of this arrangement is that the design becomes concrete. Told to build “the interest stage,” nobody knows what to make; told to build “a three-thousand-yen product,” they can. Once what to make is settled, hands move.

To rise a stage means to buy something more expensive. So the state of somebody who has not yet risen gets described as “somebody who has not yet bought the more expensive thing.” That description contains no information at all from the other person’s side.

Not-yet-purchased is an absence seen from here. On the other person’s side, something is happening. They may not yet regard it as a problem; they may regard it as a problem whose priority is low; or the priority may be high while asking this particular person has not yet entered the set of options under consideration. Three entirely different states, all of them the same “has not bought yet” on a price-ordered funnel.

Here the observation arises that all that is visible from this side is whether somebody purchased. What cannot be seen cannot serve as design input.

Where the only visible thing is the purchase record, that observation holds. Even so, things other than purchase are observable. Where they arrived from, what they read to the end, what they came back to look at a second time, which question they sent. All of these are observable at stages before purchase. They are not invisible — they are simply less countable than the purchase record.

While the scale is small, these observations do not become statistics. Calculate a drop-off proportion at a stage where a handful of people arrive per month and a different figure appears the following month. What is needed at this scale is not proportions but following each person’s path to the end. Small numbers also mean that this work is possible.

The price-ordered arrangement carries another consequence. The lower stage gets treated as the destination. The upper stages become devices for sending people downward, and what is being handed over at the upper stages stops being asked. The free document then gets built as “a document that connects to the main offer,” with contents that complete nothing on their own.

From the receiving side this is apparent. It remains as the experience of having read to the end and settled nothing. Somebody left with that experience has no reason to advance to the next stage. Thinning the upper stage in order to send people down reduces arrivals at the stage below.

That reduction has a lag. In the month the upper stage is thinned, the count passing through the upper stage may even rise. The less there is to hand over, the lower the load of reading to the end. What falls comes later. On the indicators, the causal stage improves and the resulting stage deteriorates.

What Is Arranged in the Sequence Is Stages of Involvement

What is arranged in the sequence is neither products nor measures but stages of involvement.

Involvement is the quantity of attention and time the other person is paying towards what you are saying, and simultaneously the degree to which they are treating it as something that concerns them.

Two elements sit inside it. Quantity, and whether it is their own business. The two move together but are not the same. Somebody can read at length while reading it as somebody else’s affair, and can read briefly and judge it to be about themselves.

Place involvement inside the stages and the descriptions come out like this.

• Knows this side exists but does not think it concerns them
• Thinks it concerns them but does not think it is the matter to handle now
• Thinks it is the matter to handle now but has not decided whether to ask this person
• Has decided to ask this person but the terms do not fit

All four are descriptions of a state on the other person’s side. And every one of them is equally “has not bought.” What the price-ordered arrangement collapsed into one is here separated.

Once separated, what to do changes. Showing case studies to somebody at the first stage moves nothing. Case studies work at the third. What is needed at the first is the recognition that this account has the same structure as their own problem.

Involvement cannot be measured, so making it the axis ends in guesswork. This suspicion is legitimate.

It cannot be measured directly. Even so, where the level of involvement differs, responses to the same words differ. Responses are observable. This is estimation rather than measurement, and used in the knowledge that it is estimation, it becomes something other than guesswork.

The four stages divided by involvement do not have to be these four. The division changes with what is being handled. What is fixed here is not the count of stages but the single point that the contents of a stage are written as a state on the other person’s side. Whether that has been achieved is easy to confirm. Remove your own product name from the sentence and see whether it still stands. If it does not, what is written is not the other person’s state but the name of your own offering.

Build stages out of involvement and their contents stop being settled by what is on hand. Divide the other person’s state into four and what is required is something answering each of four states, not the count of what can currently be made. Discovering that something cannot be made is also an output of the design.

Built this way, the stages sometimes increase. What took three stages in price order turns out to have five when seen through involvement. More stages mean more things to prepare.

The increase itself is not bad news. It means a stage that was invisible has become visible. Try to fill all five at once, though, and every one of them thins. In practice there is one stage where people are currently accumulating. Thickening that one alone works better than filling five evenly. Even filling comes out of seeing stages as a diagram, and a diagram draws every stage as a box of equal size.

Low Involvement Moves Perception Before It Moves Attitude

That involvement changes the landing itself was measured early in advertising research.

The psychologist Herbert Krugman studied how television advertising takes effect and held that information contacted repeatedly under low involvement rearranges the perceptual structure around the object without changing attitude (Krugman, 1965, Public Opinion Quarterly, 29(3), 349–356).

Put differently, the order reverses. Under high involvement, content is examined, attitude changes, and behaviour follows. Under low involvement, attitude stays put while the ordering of perception shifts, behaviour comes first, and attitude arrives afterwards.

What this shows is that what happens at a low-involvement stage is not “nothing.” Information reaches low-involvement recipients too. It enters differently.

So what happens when a strong claim is presented at a low-involvement stage does not split into two possibilities. The claim is not examined as content; only its position within the ordering of objects shifts. Present a different position and the position moves there.

Krugman’s observation rests heavily on the properties of television as a medium. The recipient cannot select, cannot stop midway, and the material runs repeatedly. In a medium where the reader chooses both pace and order, the conditions are not the same. That gap remains as a part which does not transfer.

The implication for design is unchanged. Try to maximise response at a low-involvement stage and the design drifts towards strengthening cues that bypass examination of content. Numbers, track records, authority, headcount. These do work on low-involvement recipients. Because they work, they continue.

What gathers as a result are the people who settled a position by that kind of cue. When a stronger cue appears elsewhere, they move there. Moving is not a failure of good faith — they were simply placed in a position where deciding works that way.

One consequence for the order of stages follows. The work of raising involvement and the work of prompting action cannot sit in the same stage. Where action occurs before involvement rises, that action is not supported by attitude. Unsupported action does not predict the next action.

Then raise involvement first — except involvement cannot be raised directly. It rises only when the account connects with the person’s own problem. The connection happens on their side, so all that can be done here is to place things in a form capable of connecting. Handing over the material for connection and prompting the connection are different jobs. The first can be written; the second cannot.

So No Number of Stages Is Common to Marketing Funnel Design

The reason the count cannot be generalised comes out from here.

How many shifts in involvement exist varies with what is being handled. In a transaction that requires somebody else’s approval before purchase, approval exists as a stage. In a small-amount transaction settled by one person, it does not. The number of stages is not a design variable to be set here but a property on the side of the object.

One thing does not vary. The ordering relation. Every stage presupposes the involvement of the one before it. Somebody who does not think this concerns them will not consider whether to ask this person. Consideration itself occurs only after the judgement that it concerns them has been settled.

In practice people skip stages. Some apply straight from a page they saw for the first time. This is a fact.

They do. What is skipped, though, are the stages prepared here, not the stages of involvement. Even somebody seeing this site for the first time has passed, somewhere, the stage of treating the problem as their own. What was passed elsewhere simply cannot be observed from here.

The distinction matters in practice. When a steady number apply straight through, the conclusion that the upper stages are unnecessary does not follow. What follows is the description that the upper stages are being supplied by somewhere else. Whether the supplier is a book, a person who referred them, or another page seen earlier — which one it is changes what to do next.

An ordering relation does not mean people advance in order. People move up and down. Somebody who reached the consideration stage returning to the interest stage because another problem took priority happens. The order is a relation of prior establishment, not a rule of progression.

How somebody who moved down is handled turns on this difference. Read as a rule of progression, they become “somebody who dropped out.” Measures aimed at recovering them get directed at anybody recorded that way. Read as a relation of prior establishment, they are “somebody at the stage above.” What somebody at the stage above needs is not recovery but whatever that stage demands.

Borrowing the count from a general template creates a second nuisance. Measures get prepared for stages that do not exist. Somebody who borrowed a four-stage template builds a measure for the fourth. In their own trade, though, the third and fourth turn out to have been one stage.

The measure for the fourth then misses. It misses, and the cause is read as the quality of the measure. The conclusion “the case studies were presented weakly” appears, and the case studies get rebuilt. What was made for a non-existent stage does not arrive however often it is rebuilt. Take the count from the side of the object and this kind of rebuilding does not occur.

The Narrowing Is Not the Aim but a Consequence of Involvement Being Finite

That the funnel diagram narrows towards the bottom admits two readings.

One reads it as filtering. Gather widely at the top and send people down while dropping those who do not fit the criteria. On this reading, narrowing is the intent of the design.

The other reads it as a consequence of there being a ceiling on the total quantity of involvement. Everybody has a limit on the attention and time they can pay in a day. Deeper stages require more involvement, and fewer people can pay the increase. On this reading, narrowing is not intent but a shape emerging automatically from a constraint on resources.

Which reading is taken changes the measures.

Read as filtering, designs that drop people become justified. Tighten the criteria, raise the price, add friction to applying. These do reduce headcount and appear to raise the quality of those remaining.

Read as a consequence of the constraint, the question changes. Who is falling away — those without the involvement to pay, or those who did not receive a reason to pay it? Two accounts stand for the same drop in headcount. Which one holds cannot be settled without following what those who fell away were looking at.

There is a trap here that is hard to see from the indicators. Operations aimed at reducing drop-off sometimes break what is downstream. Raising the pass-through rate at each stage looks correct in itself, but the easiest way to raise it is to lower the involvement that stage demands. Lower the demand and people pass through.

Those who passed do not hold the involvement the next stage demands. What was lowered above appears below. On the indicators, the upper stage improves and the lower deteriorates. And the deterioration below gets handled as a problem below. The cause is above and the treatment happens beneath it.

Lowering the demand and making something clearer are different things. If clarity raises the pass-through rate, is that not an improvement? The two are indeed different.

Different, and hard to tell apart from outside — that is where the difficulty sits. The basis for telling them apart lies after the passage. Somebody who passed because things were made clearer keeps the same quality of judgement at the stage below. Somebody who passed because the demand was lowered stalls at the stage below. Look at where they stalled and which one it was becomes apparent after the fact. Not before.

Stages that demand too much do exist. There, lowering is the correct judgement. The determination comes from looking at what the stage below was like before lowering anything. If the lower stage is healthy and only the upper is congested, the demand is too high. If both are congested, lowering the upper only worsens the lower.

This structure is taken further in where to start when the conversion rate is poor.

The Boundaries Between Stages Are Not Decided but Observed and Named

Decide the boundaries first and they become hypotheses. Being hypotheses is not itself a problem; the problem is that they stop being treated as hypotheses. The moment they are drawn on a diagram, boundaries look like facts.

Take the boundaries from observation and the order reverses. First, lay out what the people who actually came did. Then look for the places where the quality of behaviour clearly changed. The point where reading time suddenly lengthened, where the kind of question changed, where the speed of reply changed. The places that changed are the candidate boundaries.

This method has a condition: a scale at which people can be followed one by one. As numbers grow, individual following becomes impossible and the work switches to aggregation. Even so, having followed individuals once before switching means the aggregate figures carry a meaning. Without it, figures stay figures.

Boundaries taken from observation sometimes fail to match what would be convenient here. Failing to match carries more value as information. Where the number of stages prepared and the number observed differ, the content of the difference points at a gap in the design.

Deciding after observing is too slow, and at a stage where nobody has arrived nothing can be designed at all. This ordering problem does occur.

At that stage, indeed, nothing can be observed. What is needed there, though, is not the design of stages. What to do before the first person arrives is not to carve stages but to settle on one answer to the question of whose problem this account shares a structure with. Stages become necessary once the states of arrivals start to diverge.

In many cases the order is reversed. A five-stage funnel diagram gets drawn before anybody has arrived, and measures are prepared for each stage. What was prepared goes unverified because nobody comes. What accumulates unverified then works to distort observation once observation begins, because the interpretation that fits what has already been built gets selected.

What is observable is only the behaviour of those who came. Those who did not come, and those who stopped looking midway, leave no behavioural record. Boundaries taken from observation are therefore boundaries as seen from the side of those who stayed. That bias does not disappear. It has to be used in the knowledge that it does not.

And the bias strengthens over time. Keep taking boundaries from the side of those who stayed and the stages tidy themselves year by year into whatever suits this side. Those who stayed are the ones who adapted to how things are presented here, so boundaries taken from their behaviour end up ratifying that presentation.

The material for breaking this exists only outside the observation. Those who did not come, those who stopped midway, those handling the same problem elsewhere. Periodically, the side absent from your records has to be looked at. Keep updating a design from what remains in the records alone and the field of view narrows with each update. That it narrowed does not appear in the records.

Does Each Stage Leave Something Behind for Somebody Who Steps Off There

A common failure in design is a state where the contents of a stage are built solely to send people to the next one.

From the building side, this form looks rational. A funnel is a device for sending people downward, so each stage need only perform the sending — that is the reasoning.

The reasoning collides with involvement as a variable. Involvement rises as a result of receiving something. Somebody who received contents built only for sending does not rise in involvement. Not having risen, they are sent onward without reaching the level the next stage demands.

Hand everything over at the upper stage and there is nothing left to sell below. What this worry sets up is a way of counting what is handed over as quantity.

Framed as a question of quantity, no answer emerges. That is why the question of how much to give away free gets debated repeatedly without settling.

The dividing line is not quantity but the gap between a principle and an individual answer. Hand over the structure of why something happens and the work of applying it to that person’s situation remains. The applying is what is being handed over. Conversely, withhold the principle and hand over only the answer tailored to them, and the transaction ends the moment that answer has been delivered.

How this line is drawn is also treated in where the value of an intangible product resides.

Where each stage completes something, more people step off. Their business was done there. On the indicators, the pass-through rate appears to have fallen.

Those who stepped off, though, did not step off dissatisfied. Where somebody who remembers having their business done goes when they hold a different problem. The pass-through rate cannot record that time lag. Judge by recordable indicators alone something whose nature is unrecordable, and the conclusion emerges that not completing anything is correct.

Not every stage can be made to complete something. At some stages, what can be handed over is intrinsically small. Stages for confirming terms or arranging schedules hand over nothing in themselves.

A stage with nothing to hand over and a stage that could hand something over and does not are separate things. The first is a formal stage, so shortening it is correct. The second is a stage whose contents were hollowed out because the aim of sending people down came first. From outside, both look like “thin stages.” Telling them apart means asking whether the content that stage handles could have independent value at all.

Where each stage completes something, that stage starts being referenced on its own. People begin arriving at that stage alone from elsewhere. More arrivals have not come in order from the top of the funnel.

As a diagram, this state looks broken. What is actually happening is that entrances have increased. That entrances increased is itself welcome, but unless it is settled where people from each entrance converge, the added entrances become exits. This problem is treated separately in customer journey design.

Read the Questions Coming from Below, Not the Pass-Through Rate

Whether a stage exists cannot be confirmed from the count passing through. Look at the state of those who passed.

The first is the quality of the questions coming from the stage below. If a question from somebody at the consideration stage is a confirmation of something already handled above, the upper stage did not arrive. If it is a question that builds on the upper stage, involvement rose as the stages describe. The same count may pass, and where the questions differ, what happened differs.

The second: does the other person use vocabulary you did not supply? Somebody who translated what they received into their own situation talks about it in words you do not use. Where no translation occurred, your own words come back unchanged.

The third: can the origin be identified for somebody who skipped stages? That people skip is not itself the problem. The problem is a state where it is unknown where they passed the upper stages. Unknown means the supply cannot be reproduced. Traffic that cannot be reproduced is not something whose continuation is settled on this side.

None of the three looks at the count who passed. All of them look at the state of those who passed. Since a funnel is a description of stages of involvement, what has to be confirmed is involvement.

The reason the pass-through rate sits outside the principal indicators is here too. The pass-through rate turns two different events into the same figure. Somebody who passed because involvement rose stage by stage and somebody who passed because the demand was lowered are counted as the same single case. Two routes make the figure improve, and one of them breaks what is downstream. Which route produced the improvement cannot be told from the pass-through rate itself.

Seeing states requires reading one person at a time, so it will not hold once headcount grows. That worry mistakes how the three indicators are used.

Reading every case is unnecessary. All three are of a kind that can be judged from a small number of instances. The quality of questions, the vocabulary in use, the origin of traffic — read around ten and a first sign of the tendency emerges. This is the work of seeing kinds, not of producing proportions.

The reason no number is supplied here is not that none could be. How many stages, how much to hand over at each, which indicator to watch principally — the meaning of these three inverts the moment a number is attached. Write “four stages” and people appear who borrow the four-stage template. The borrowed template then matches, again, what they have on hand. That returns to the question of how many stages to prepare.

What has been handled is not how many stages to have but why nobody can answer that question head on. The reason is that the number of stages is a property on the side of the object. What sits on the side of the object does not emerge without looking at the object.

Look at your own diagram, and the first thing confirmable. Delete every product name and service name of your own from the description of each stage. Any stage whose sentence still stands is written as a state on the other person’s side. Any stage where nothing remains was your own inventory arranged as stages. The stages that left nothing behind are the ones that do not currently exist.