In Customer Journey Design, Hold Several Entrances and Narrow the Confluence to One, and a History Remains

Customer journey design settles two things and no more. Where people are allowed to enter, and where those who entered converge into one. The two differ entirely in the nature of their ceilings.

There is no ceiling on the number of entrances. There may be as many as there are places where the other person already is, and more of them widens the range that can be reached. Whether to add more is settled by how much you can produce and how much administration you can carry. It is a variable you get to choose.

Only one confluence of a given role can be held. With two or more, what ought to happen there stops happening. It is not a variable to choose but a structural constraint. Things of different roles standing side by side — an enquiry window and a continuing contact line — do not fall under this.

This asymmetry is the spine of customer journey design. In the circulating accounts, though, both sit under the same word. Because they share a name, the judgement to narrow entrances and the judgement to split the confluence both look like the same adjustment. Over-narrow one or over-split the other and, on screen, the work has a similar shape.

Three things are settled here. Where the entrances are. Why the confluence has to be one. And how to confirm that the route is identifiable.

Some things do not get resolved. How many entrances you ought to hold is not settled here. That value is determined by where the other person already is, and that distribution exists only in your field. Nor is how to vary what gets delivered. That belongs to the design of reader state, not to the design of the receptacle.

With two or more confluences, somebody who made contact five times becomes, in the records, two different people who made contact twice and three times. And split records cannot be merged afterwards. The common key for merging them was never held on either side.

The split shows up on the indicators as good news, because two new registrations appear. A state where one person looks like two and a state where two people actually came are indistinguishable in the records. It becomes noticeable only when the response to what was delivered fails to match the growth in registrations for a sustained period. By then the split has accumulated for some months.

And whether a split is happening now can be counted from outside. When you put something out, does the same person respond in more than one place? If they do, that person exists more than once in your records. If they do not, one of the two is simply unused. Both come out of holding two receptacles.

The Customer Journey Design in Circulation Has a Boundary Drawn Around What It Covers

The customer journey design that gets presented is accurate as practice. Its range, though, is cut off by a single line.

Customer journey design is the design for guiding somebody who visits a site smoothly to a target page. The target page usually means the completion screen for a purchase, an enquiry, a document request, or a registration.

A distinction from the observed route is usually attached. The designed route is the path prepared here. The observed route is the track people actually took. The two do not coincide. On that premise, the procedure presented is to rebuild the designed route while watching the record of the observed one.

The specific measures listed are the placement of calls to action within a page, the depth of navigation, the presentation of related articles, the reduction of input fields, and page load speed. All of them reduce the friction on the way to the target page.

The effects cited are improvements in circulation and conversion. Where operational records have accumulated, the advice adds that designing from the observed route is effective.

This far is accurate. Reduce friction and arrivals increase. Build the designed route without looking at the observed one and it gets built from assumptions here alone.

Looking at the whole set, the boundary sits in a single place. Every measure draws its line at the perimeter of the site. The frame is: how to move somebody who has visited. Before the visit and after the departure lie outside it.

The frame is not invalid. In a business where visits are already sufficient and only the friction before arrival is at issue, work inside the frame works best.

It breaks when the frame is started from while visits themselves are unstable. Inside the frame there is neither an operation for increasing entrances nor one for settling the confluence. What is not inside does not become a candidate for consideration. The result is that only the internal friction gets adjusted, repeatedly.

Starting from inside is correct as an ordering — that counter-argument stands. Fixing what is at hand is what practice looks like.

Fixing what is at hand is itself a correct judgement. What “at hand” means here, though, is the range that can be fixed, not the range where the cause lies. They are different things, and while the work is under way there is no telling them apart. In either case the hands move identically.

And internal adjustment usually comes with some change in the figures. Move a button and the proportion of presses changes. The changed figure gets recorded as the result of the work.

That records remain is the reason this work continues. Increasing entrances and settling the confluence, meanwhile, produce no figures in the month they are started. Keep selecting whichever produces short-term figures and the other side is never started at all. The criterion of selection has become not effect but visibility of effect.

What the Distinction Between the Designed Route and the Observed Route Conceals

The distinction is useful, and both terms stand on the same side.

It is useful because it names, separately, what was intended here and what actually happened. Without the naming, when people fail to travel the intended path, the failure itself goes unobserved.

The distinction has one property that is hard to see. Both are defined inside the site.

The designed route is described as connections between pages, and the observed route is recorded as transitions between pages. Both concern what happens after arrival. Where the person was before arriving, and where they went after leaving, cannot be handled in this vocabulary.

What the pair can speak about is only the middle of the route. The start and the end lie outside the pair.

The start and the end cannot be measured, so of course they cannot be handled — this observation is half right.

What cannot be measured and what the vocabulary does not hold are different things. The start can be measured. Where somebody arrived from is recorded in most cases. The end, too, is partly recorded in the form of what they next came to look at.

The reason the measurable goes unhandled is that neither falls within the range the word “route” points at. What sits outside the vocabulary does not rise as a candidate for improvement.

Move the boundary and the object of design changes. Not the perimeter of the site, but from the place where the person first registered that you exist, through to the state of being reachable continuously — treated as one continuous route. Seen across that range, the middle friction becomes one part of a route.

Widen the range and the variables multiply beyond control, a concern which stands. In fact the reverse holds. Across the wider range there are only two places that matter most: the number of entrances and the position of the confluence.

The middle friction has an unlimited supply of fine adjustments. Colour, wording, position, order. Entrances and confluences, by contrast, are discrete. You hold one or you do not. Widen the range of what is handled and the number of judgements falls.

That fall comes at a cost. The weight of each single judgement rises. A button can be put back next week if it was wrong; move the confluence midway and the history accumulated until then becomes unusable. Judgements that cannot be reversed are the ones that have to be settled first.

The weight also works to delay the start. Judgements known to be irreversible get held until the material is ready. The material, though, only gathers after the decision. People arriving during the period when the confluence is unsettled pass through without accumulating anywhere. The length of the holding period is the volume of history lost.

Entrances Are Not Made Here; They Already Exist

An entrance is the point where a place the other person already occupies makes contact with your content.

The point of this definition is that the position of an entrance is not settled by what suits you. People are not travelling towards your site. They are somewhere handling their own problem, and from there they make contact with you by chance. Where contact occurs is determined by where they are.

Increasing entrances therefore means increasing the kinds of places the other person occupies. Search, books, referrals, references from other articles, mentions by others. All of these are different places and each is a separate entrance.

More entrances make administration harder; narrowing to one and thickening it is more efficient. On administrative load, that holds.

Narrowing carries a property beyond efficiency, though. With one entrance, when that one becomes unusable, arrivals go to zero.

This is not a hypothetical. Search rankings move by somebody else’s decision. The side that was referencing you may delete the article. Where the maintenance of an entrance does not conclude on your side, that entrance is a borrowed route. Borrowed routes vanish when the lender changes terms.

Adding more is not simply better either. The number you can add is bounded by how much you can produce. Exceed the bound and every entrance thins.

The determination is not the count but whether each entrance functions independently. If somebody arriving through an entrance has their business done by that entrance’s contents alone, that entrance is independent. Where it presupposes the contents of another entrance, there is in fact one entrance wearing the appearance of several.

Add entrances and the states of arrivals scatter. Somebody from search and somebody from a referral differ in their level of involvement towards you from the outset. Present the same thing and it is too strong for one and insufficient for the other. That the level of involvement differs by entrance bears directly on how the stages get cut. Cutting stages on involvement is treated in the design of a marketing funnel.

The scatter is not a side consequence of adding entrances; it existed before they were added. When there was one entrance, the scattered people simply were not arriving. Adding them is what first made a response necessary.

Mistake that order and the treatment reverses. Read the scatter as “harm caused by adding” and the direction turns back towards fewer entrances. Turn back and the scatter subsides and the indicators tidy up. Tidy indicators do not record the existence of those who are not arriving.

The amount of explanation required afterwards differs by entrance too. Somebody from search does not know who you are. Somebody from a referral holds one judgement made by the referrer. Show them the same page and one reads from the premises while the other skips the premises and reads from the substance.

Thicken the explanation of premises for everybody and it becomes redundant for the referred. Thin it and it is insufficient for those from search. Framed as a choice between the two, it does not resolve. It resolves when the explanation of premises is moved out of the body text into a form that only those who need it step onto. The more entrances there are, the more this separation is demanded.

Where There Are Several Confluences, the History of Involvement Remains Nowhere

A confluence is the place where everybody, whatever entrance they came through, finally gathers into one. Gathering means, from your side, entering a state where you can deliver to them continuously.

The reason there must be one is that involvement has the property of accumulating.

When somebody who read once reads a second time, the second is stacked on the first. The third is stacked on the second. That accumulation raises the stages of involvement. Accumulation is recorded, however, only where you can recognise the person as the same person.

With two confluences, the same person sometimes registers at both. Or registers at only one. In either case that person’s history of involvement is split. Split histories appear, on each side, as “a new person.”

Somebody reached five times splits, in the records, into two people. That is what happens where confluences are plural.

Even with several confluences, more options make registering easier; narrow the window and those who do not fit fall away. This worry has a body of experiments addressing it.

In Sheena Iyengar and Mark Lepper’s 2000 study, the condition with more options drew more people to stop by, while the proportion who went on to purchase was lower (Iyengar & Lepper, 2000, Journal of Personality and Social Psychology, 79(6), 995–1006). That finding cannot be used as a general rule, though. In a meta-analysis by Benjamin Scheibehenne and colleagues covering fifty experiments, the mean effect size was near zero with large variation between studies (Scheibehenne, Greifeneder & Todd, 2010, Journal of Consumer Research, 37(3), 409–425).

So “too many options and people cannot decide” is weak ground for making the confluence single. Depending on conditions, it happens and it does not.

The reason for a single confluence is not the load of choosing. It is whether history accumulates. There are settings where many options do not slow the decision, but once a history is split, the fact of the split holds regardless of conditions. Place the reason in the load of choosing and the reason disappears with the next experimental result. Place it in accumulation and it does not.

Several confluences of different kinds and duplicated confluences of the same kind are separate matters. An enquiry window and a continuing contact line have different roles. Things with different roles may be plural. Only where things of the same role are plural does the history split.

Split histories cannot be merged later because there is no key. Even where the same person is registered in two receptacles, they can be judged the same only where both use the same identifier. Without one, the judgement becomes a comparison of names and registration dates, which is not proof of identity.

The Confluence Has to Be a Place Whose Terms Are Settled on This Side

Beyond singularity, a confluence has one further condition. The terms of that place have to be settled on your side.

Terms means the range of people reached, the frequency of reaching them, the form of delivery, and whether the place can continue to be used.

A place whose terms sit on somebody else’s side does not function as a confluence. Not functioning does not mean nothing arrives. Some days it arrives. It means you cannot settle which days it arrives and which it does not.

Concretely it appears like this. You put out the same content and the number reached varies widely by the day. No reason is given, and it cannot be reproduced on your side. What cannot be reproduced cannot serve as an element of design. Build the unusable into a design and you repeat the same actions next month without knowing why the good month was good.

Terms settled on your side do not guarantee arrival. Whether something is received where it is sent depends on the recipient’s settings too. Owning and always arriving are different things.

What separates them is what becomes knowable when something fails to arrive. Where terms are settled on your side, the count and the reason for failure are recorded. Once recorded, the next move is settled. In a borrowed place, the failure itself goes unobserved. All that remains is the fact that something was put out and nothing came back, with no way to separate not-delivered from delivered-but-unread.

Without that separation, the estimate of cause always points inward. The content must have been poor; the timing must have been bad. Where estimates only ever point inward, improvement gets fixed to rebuilding content. The same conclusion emerges even where the stoppage is on the route.

And that fixing continues quietly. If the rebuilt content is better than before, the response moves somewhat. The movement gets read as evidence that the estimate was right. The route never comes up as a candidate.

Terms sitting on somebody else’s side do not necessarily show as a loss today. The periods when the terms work in your favour may be longer. The problem is that the switch between favourable and unfavourable cannot be predicted from here. Build a design on top of an unpredictable switch and the loss at the moment of switching grows in proportion to what was built.

The determination therefore cannot be made from “is it working now.” It is made from the single point of who settles the terms. Judge during a favourable period and who settles them is invisible. What is visible is the result, and for the time being the result has the same shape either way.

Movement from Entrance to Confluence Is the Single Point of Customer Journey Design

Not using places where many people already are means discarding reach. This reading does not contain the answer “do not use them.”

It is the matter of separating using something as an entrance from using it as a confluence. Places with many people are excellent as entrances, because the other person is already there.

Separate them, then have people move from the entrance to the confluence. That movement is the single most important place in customer journey design. Where the movement does not occur, involvement does not accumulate however wide the entrances are.

There is a practical ceiling on how often movement may be requested. Ask every time and the contents of that entrance start to look like a notice about moving. Once the contents look like a notice, that entrance stops functioning as an entrance too.

A ceiling means the position has to be chosen. The basis for choosing is not frequency. It is whether the content did somebody’s business on the spot. Immediately after their business is done, the reason to move stands on their side. Ask while their business is undone and the movement gets processed as “asked for the next thing before receiving anything.”

Put the confluence on your side and the early figures fall. Borrowed places produce more response in situ. Insert a movement and the count necessarily drops.

The lost portion has shifted to the side that stays on record. The smaller number accumulates and the larger does not. Confirming that inversion in the figures takes time, so judging early tends to make the borrowed side look correct.

And the pull towards judging early is strong. Figures from a borrowed place appear the same day. The accumulation on the other side starts to carry meaning months later. The two can be set side by side only where the period of comparison was fixed beforehand.

Without a fixed period, the comparison always happens at “today’s figures.” Today’s figures always favour the borrowed side. The error therefore never appears as a matter of capability, only as a matter of when the comparison was made.

The design of movement carries one further practical condition. What happens after the movement has to be shown at the moment the movement is requested. Ask without showing and the other person infers what follows. The inference usually lands on “a sales pitch is coming.” Once it lands there, the movement does not occur.

Showing does not require announcing frequency or contents. It is enough that what lies past the movement is evidently of the same nature as what is being read now. If the nature differs, the movement becomes the start of a different relationship. Starting a different relationship requires something the present one lacks.

Do Not Build a Point Demanding a Judgement Midway Along the Route

Where a point demanding a judgement sits midway along a route, people stop there. Not because the judgement is difficult.

Moving and judging are different acts. Moving is a change of position: advancing to the next page, following a link. Judging is deciding which to select, or whether to decide now.

Judging requires a different kind of resource from moving. Whether that resource can be spent right now is not visible from this side.

Judgements therefore go at the ends of a route. The middle carries movement only. A judgement placed at an end is made by somebody who has finished moving.

Concretely it takes this form. Do not put a fork midway through an article asking which course suits the reader. Put one fork after the reading is done. A fork placed midway demands a selection in place of continued reading.

Collecting judgements at the ends is not the same as reducing them. Continue reducing and the route ends up with no judgement at all. Somebody who travelled a route with no judgement arrives at the confluence having decided nothing. Where somebody arrives undecided, accumulation does not begin.

Once the position of the judgement is settled, what has to be handed over before it is settled too. If the judgement at the end is “whether to keep reading this person,” what has to be handed over beforehand is the material for judging that. The material has to be something already established, not a summary of your claims.

Confirming intent midway lets you present something better suited to the person; the fork is for their sake. There are cases where it does.

The fork midway, though, demands a judgement at a stage where the material for judging is not yet assembled. Which suits you is not settled without knowing the contents of both. Made to choose before knowing, what gets chosen is whichever name is easier to grasp.

After choosing, only the contents of the chosen side arrive. The first fork fixes the range of what can be received thereafter. The earlier the fork, the wider the range fixed.

Collect judgements at the ends and nothing is known about those who left midway. With a fork midway, where and what they selected is recorded. Remove the fork and the record goes too. The quality of the design and the granularity of observation point in opposite directions here. The judgement to add forks in order to add records overlooks that opposition.

And the oversight cannot be detected from the indicators. Add forks and the number of recordable items grows. More items mean more that can be analysed. The sense that the analysis has grown richer covers over the fact that the route has been cut into pieces.

Build a Receptacle per Entrance and the Ones You Cannot Fold Away Remain

One common breakage is building a dedicated receptacle alongside each entrance that gets added.

At the moment of building, the form looks rational. Guidance for somebody from search and guidance for somebody from a referral should of course differ in content. Separating the content is correct as far as it goes. What must not be separated is the receptacle beyond it.

Separate the receptacles too and this follows. People accumulate in only one of them while the other is maintained empty. Try to fold away the empty one and the question arises of what to do with the handful registered there. Being unable to fold it, both remain.

Then people registered in both appear. The same content arrives twice, so one gets cancelled. The cancellation is recorded on your side as a departure. Nobody departed. A duplicate was resolved. The record cannot distinguish the two.

If the content is to differ by entrance, the receptacle has to be separated too. It does not.

Make the receptacle single and hold an attribute inside it. Keep a record of which entrance somebody came through and vary what gets delivered from there. The number of receptacles and the number of kinds of content are independent variables.

The detail of this division is treated in segmented email delivery. What has to be confirmed here is the single point that “I want to vary the content” is not a reason for “separate the receptacle.”

Where the fields being handled genuinely are separate, separating is sometimes correct. The determination comes from whether people registered in both appear. If they do, on the other person’s side those two are one interest. If they do not, they are separate interests.

A state with more receptacles, left alone, fixes as much operational labour as it added. A receptacle once built has registrants, even a handful. The judgement to stop a place with registrants gets harder to make the fewer they are. The impact should be small because they are few, and it is precisely because they are few that the question takes the form “are you cutting these people off?”

The result is that the receptacles that function least survive longest. While they survive, something has to be issued into them periodically. Otherwise it is a place where nothing arrives for those who registered. The operational load falls disproportionately on the non-functioning side.

There is only one way to avoid this structure. Before adding a receptacle, settle how it will be folded away. A receptacle with no settled way of folding is effectively built as permanent. Knowing it is permanent changes the judgement of whether to add it at all.

Not Circulation, but Whether a Second Visit Is Visible as a Second Visit

Whether a route is identifiable cannot be read from circulation or pages-to-arrival.

The first is whether the entrance can be identified for somebody who reached the confluence. Where it can, that entrance is reproducible. Where much of the traffic is unidentified, what has increased is arrivals, not design. Whether the same number appears next month cannot be predicted from your side.

The second is whether a second visit by the same person is recognisable as a second visit. If it is not, involvement is not accumulating. In that state, every contact is a first contact however many there are. Measures that increase contact frequency have no effect in this state.

The third is whether a steady number have their business done by an entrance’s contents alone. A state where everybody proceeds to the confluence is not a good state. Either the entrance is not functioning independently, or the option of not proceeding has not really been prepared. That some people have their business done and step off is evidence that the entrance holds substance.

None of the three looks at the count who passed; all three look at whether the route is identifiable. Since customer journey design is the design of entrances and confluence, what has to be confirmed is the route.

The reason circulation sits outside the principal indicators is here too. Circulation turns two different states into the same figure. Somebody who read on because their interest continued and somebody who moved to another page because they could not find what they sought are both counted as circulation. The latter is a movement that ought to be detected as a target for improvement. Far from being detected, it enters on the side that improves the indicator.

Identifying entrances and recognising individuals both require a measurement apparatus. While the scale is small, no apparatus is required.

Ask the people who came and it becomes clear. Where did you hear about this, have you read something here before. Ask around ten and a first sign of the tendency emerges. An apparatus becomes necessary once the count exceeds what can be asked.

And whether you have the memory of the period when you asked changes how the figures read after the apparatus arrives. To somebody who never asked, aggregated figures are figures from the start. Having asked even once, what stands behind those figures is legible.

How the confluence itself is held — what has to be settled before “collecting” — is treated in what has to be settled before list building.

The right number of entrances is not on your side. It is a value determined by where the other person is, which makes the distribution the thing that decides. The position for requesting movement and the attributes to record have the same shape. The first is settled by where the content did somebody’s business; the second by the unit you will later want to divide along. In every case, what settles them is what you are looking at. Put a number first and what you are looking at gets replaced by the number.

The side with no ceiling and the side whose ceiling is one. As long as those two are mistaken for each other, no amount of adjustment inside makes the route identifiable.

So what can be confirmed now splits in two. If the confluence is single, what to look at next is the number of entrances. If they are insufficient, there is room to add, and room to add works before any middle friction. If there are two or more confluences, the number of entrances waits. What comes first is deciding which to fold away. That judgement gets harder as registrants grow. If there are two now, now is the easiest moment there will be.