Customer Journey Mapping: Map the Decisions, Not the Touchpoints, and End With Fixes

Customer journey mapping is the practice of drawing what a defined customer tries to do, decides and feels from the first reason to act to the outcome they wanted, then using the gaps to choose fixes. A journey map is useful only if it ends in changes with owners, deadlines and metrics.

Nielsen Norman Group, the UX research firm, lists five parts of a journey map: an actor, a scenario with the actor's expectations, journey phases, the actions, mindsets and emotions inside each phase, and the opportunities the mapping reveals. It names two benefits. Building a map forces a team to share one mental model, and the finished map is a compact way to pass that picture to people who were not in the room.

Notice what that list leaves out. None of the five parts says who will act on an opportunity, or by when. The same firm advises one point of view per map and a narrative of the steps, not a log of every interaction, and says maps suit scenarios that involve a sequence of events, a process or several channels. So the technique fits a sales process or a booking flow well and fits 'the whole customer experience' badly.

This guide keeps the technique and changes its unit. Instead of plotting touchpoints (the ad, the email, the call), it plots the decisions the person makes between them, and it attaches a fix register so that the output is a list of changes rather than a poster.

Nielsen Norman Group surveyed 48 UX professionals about how their organizations use journey mapping. That is a small, self-selected group, so treat the figures as texture, not as industry rates. The pattern is still instructive. Asked when mapping fails, the most common answer, at 36%, was that nobody defined the purpose up front. A quarter said the map was not based on reality, and 21% said it was never shared or used.

Asked what makes mapping succeed, 27% pointed to the map leading to action. The report also notes that recommendations, metrics and ownership were the elements practitioners included least often. That is the whole problem in one line: the parts that turn a picture into work are the parts most often left off.

Our reading is that maps rarely fail at the drawing. They fail at the handover. A workshop produces a large, satisfying artifact, the room agrees, and every opportunity then belongs to everyone, which means it belongs to no one. The fix register described below exists to prevent that, and it is deliberately dull.

A touchpoint map asks where the customer met you. A decision map asks what the customer was trying to settle at that moment and what would have let them settle it in your favor. The second question is the useful one because it points at something you can change: an answer that was missing, a doubt left unaddressed, a step that asked for too much too early.

Two research findings shape how to draw it. Google and The Behavioural Architects, studying online shoppers, described people moving between states of exploration and evaluation instead of marching down a funnel. That research covers consumer purchases, so it will not describe every business sale, but it supports a design choice: treat phases as states a person can re-enter, and expect the same decision to come up more than once.

The second finding matters for B2B. Professor John Dawes of the Ehrenberg-Bass Institute, working with LinkedIn's B2B Institute, puts the share of business buyers who are in the market at any moment at about 5%. If that holds in your category, a map that starts at the first search skips the longer stretch in which preference formed. We would add a row for it and mark it as an assumption until your own buyers tell you otherwise.

Keep the table above for the diagnosis, then build a second table: the fix register. Each gap becomes one row, and a row is not finished until every column is filled. Some columns will feel bureaucratic. Each exists because of a specific way maps die: no owner, no date, no way to tell whether the fix worked, and no condition under which the team admits it did not.

Two judgment calls are ours rather than research findings. First, we suggest keeping the register to about five fixes per map per quarter. A map can easily surface twenty gaps, and a team that tries twenty fixes finishes none; the rest wait in a parked list that gets reread at the review. Second, choose the metric at the decision, not at the end of the funnel. If the gap is 'no arrival window in the first reply', the useful measure is how many first replies include one, and what share of those prospects book. The sale at the end of the funnel is moved by too many other things to credit a single fix.

Give every fix a review date within ninety days and a kill condition: the result at which you stop or reverse it. Writing the kill condition in advance is awkward, which is why it works.

The second most common failure in the survey was a map not based on reality, and the cure is to grade each cell. 'Observed' means you saw the behavior in data you can open: analytics paths, reply timestamps, recordings, form abandonment. 'Reported' means someone told you: interviews, support tickets, sales notes, survey answers. 'Assumed' means the team believes it and has not checked. Most first-draft maps are mostly assumed, which is fine as long as the map says so.

Be wary of reconstructing journeys from analytics alone. Tools bound what they can credit: in Google Analytics, for instance, the lookback window for most key events defaults to 90 days and for acquisition events to 30 days, so a decision that started earlier than that is invisible to the path report. Cross-device and offline steps, such as a phone call after reading a page or a recommendation at a dinner, usually leave no trace at all. That is a reason to interview customers, not a reason to skip the data.

A decision map does its best work when it points at something concrete. Two of our existing pieces take single decisions in depth. Our guide on why the booking button is part of the local customer journey follows what happens when a person taps through to book, and our B2B website design guide, which starts from the questions buyers bring, turns the shortlist and trust decisions into page structure.

Use the map to decide which of those pieces of work to do first. If the largest gap is at the contact decision, the fix is probably in a link, a number or a form. If it sits earlier, at the shortlist, no page change will help and the fix lives in your listing, your reviews or your reputation among the people who recommend you.

Be ready to skip the workshop. A map is the wrong tool when the problem is already known: if a form returns an error on one phone model, fix it and move on. It is also the wrong tool at very low volume. If you sell to a few dozen customers a year, ten conversations will tell you more than a diagram. And it is wasted when the people in the room cannot change the steps the map exposes, and that happens easily when ads, pages and sales follow-up sit with three different people.

Maps also fail from scope. The most common pattern in the survey's failure stories was no defined focus, and a map that tries to cover acquisition, onboarding and renewal for four customer types is really four maps. Cut it. Another quiet failure is the one-time map: a customer's options change when a competitor moves or a platform changes its rules, and a map older than a year is a record of what used to be true.

Hang the fix register, not the map, where the team will see it, and open every monthly meeting with it. The map is a reference; the register is the plan. At each review, change the evidence grade of any cell you have learned something about. A map that slowly turns from assumed to observed is getting better, and one that never changes grade is not being read.

Revisit the whole map when something structural changes: a new channel, a pricing change, a new competitor, a new location. Otherwise, run a thirty-minute check each quarter on three things: which fixes shipped, which metric moved, and which assumption you can now retire. If the answer to the first is none, the problem is ownership, and no amount of redrawing will fix it.