Customer journey research: find where a plausible purchase stops

A customer likes the offer, tries the experience and reaches checkout. Then the purchase stops. Marketing wants a better reminder email. Product wants fewer checkout steps. Operations says customers need to understand the service before joining.

Each team can point to part of the journey. The decision is which explanation deserves the next change.

Customer journey research follows an attempt to buy or use something across the encounters that shape it. It connects recorded actions with the customer's account, the alternatives available and the conditions the business can actually meet. The useful result is a specific intervention, an owner and observable behavior that would show whether it helped.

Consider Lantern, a fictional fitness studio selling memberships to people who want regular classes after work. The company, people, records and quotations throughout this walkthrough are invented. Lantern needs to decide whether to improve enrollment or change the offer behind it.

Start with the purchase the customer is trying to make

Maya wants to attend a Monday evening class with a friend. She discovers Lantern through an advert, enjoys a trial session and opens the membership checkout. The dashboard records an unfinished enrollment.

“Reduce checkout abandonment” describes the business's symptom. Maya's task is more specific: establish a routine she can actually attend. A membership is useful only if a suitable class is available at the required time.

That distinction sets the research boundary. Begin before checkout, when Maya forms an expectation about the schedule. Continue until she joins, chooses another option or stops looking. Include the studio visit, staff conversation, booking system and friend who affects the choice.

This follows the principle behind mapping a user's whole problem: investigate how online encounters, offline encounters and the processes behind them join up. A map organized only around departments can miss the handoff that matters to the customer.

Our customer needs analysis guide helps distinguish the desired outcome from a requested feature. Here, “show me a better calendar” may be a proposed solution. A dependable Monday routine is the need it must serve.

Build one connected account of what happened

Reconstruct Maya's attempt before summarizing it as a theme such as convenience or trust.

Keep the source and its limits attached to each entry. An attendance record establishes the trial visit. A later interview supplies Maya's recollection of the conversation. It does not produce a recording of what the instructor said.

Connect the records using evidence you can justify. Maya's permission to examine her own confirmation and messages can support this reconstruction. A mobile session and an unrelated laptop visit should not become one person's journey because their timing looks plausible. Leave unmatched encounters unmatched.

Check the outcome in the business's transaction records as well as its analytics. A missing tracked purchase event alone cannot establish that nobody bought. Equally, a checkout visit is not evidence that a membership was purchased.

Separate the event from the explanation

Maya says: “I liked the class. I couldn't tell whether paying would get me a place after work.” In the interview, she describes choosing another studio with a bookable weekly place that fits her schedule. Lantern has her account of that choice, rather than the other studio's transaction record.

That account gives the team a plausible explanation: Lantern communicated permission to book when Maya needed confidence that a usable place existed. It does not establish that every unfinished enrollment has the same cause.

Three possibilities lead to different work:

• Misunderstanding: a suitable place exists, but the customer cannot discover or interpret it. Clearer availability information may help.
• An unmet condition: the required class is full. Better wording cannot create a place. The schedule or membership offer needs attention.
• An experience failure: a suitable place is available and understood, but enrollment or booking fails. Reproducing the failure matters more than another message test.

Inspect the booking history and eligibility rules for the dates in the episode. “Monday is full today” cannot establish what Maya could book last week. Ask the support team to interpret its reply, and check whether the same answer appears on the website and at the studio.

Keep other explanations alive. The membership commitment, the friend's availability or a payment problem may also matter. Maya's account helps choose what to investigate; it does not isolate a causal effect by itself.

Lantern then checks the dated booking history. The Monday class had no available places when Maya enquired, and membership did not reserve one. That establishes an unmet condition in this episode. The support reply explained access without resolving it. A shorter checkout could not supply the place Maya needed.

Find the case that challenges your preferred story

Sam sees the same waitlisted Monday class and joins anyway. He works flexible hours and can attend on Tuesday morning. He understands that membership gives booking access rather than a reserved weekly place.

The waitlist has not disappeared. Its significance changes with the customer's alternatives.

This contrary case narrows the problem. Lantern should investigate people who need a particular session separately from people who can choose among several sessions. Calling both groups “busy professionals” would hide the constraint that changes the decision.

It also changes the proposed fix. A general reassurance that there are plenty of classes could disappoint Maya while accurately describing Sam's options. Useful information shows the times that are available to that customer and the commitment the offer actually supports.

Seek other contrasts deliberately: someone who bought despite uncertainty, someone who understood the rules and left, and someone who could not complete an otherwise suitable purchase. Choose these accounts to test explanations, not to estimate their prevalence from a small selected sample.

Choose research that can distinguish the causes

Choose the next study around the unresolved question. More interviews with enthusiastic trial attendees may produce praise without explaining enrollment failures.

Start with recent attempts from people who joined, delayed and chose an alternative. Ask them to reconstruct the sequence using the material they actually encountered. “What happened after you looked at the schedule?” is more useful here than “Would clearer information make you join?”

Our qualitative market research guide explains how to preserve the episode and examine competing interpretations. Where a task needs checking, watch someone try it with their usual device and relevant circumstances. Contextual observation can reveal barriers and workarounds in the setting where the activity happens. An explanation given during that session still needs to be distinguished from the action observed.

Then check operational feasibility. For Lantern, that means booking rules, available places, waiting-list behavior and what staff can promise. If everyone who needs Monday evening faces a full class, a new enrollment message addresses the wrong problem.

Use a broader measurement study when the decision depends on how common each constraint is. Define the population and recruit beyond the easiest customers to reach. The proportion of interviewed people mentioning a waitlist describes those interviews; it is not automatically the share of all potential members blocked by capacity.

Give the intervention an owner

The journey investigation should leave Lantern with a decision about what to change first.

For Maya's episode, Lantern should make availability clear before payment and stop implying that membership alone supplies her Monday routine. Operations must decide whether a dependable place can be offered. The digital team owns how customers discover the actual options. Neither can complete the fix alone.

If no place exists, Lantern must decide whether it can provide one or should direct that customer elsewhere. A successful journey can include helping someone recognize a poor fit before buying. Counting every additional enrollment as improvement would reward a promise the studio cannot deliver.

Retest the behavior the change is meant to enable

Suppose Lantern introduces an availability view before membership enrollment. The hypothesis is that people with a usable session can identify it and proceed with a clearer understanding of what they are buying.

First test comprehension and the task itself. Can the person identify an available class, explain whether the place is reserved and complete the relevant booking? Correct answers in a research session support that limited finding. They do not establish a membership increase.

For a live evaluation, follow eligible customers through membership purchase, a suitable booking and attendance. Set the follow-up period around the normal decision and attendance cycle before reading the result. Keep people who lack a suitable session visible as a separate group rather than removing them afterward to improve the conversion figure.

Check consequences beyond the next click: refunds, early cancellations, booking-related support and whether existing members lose access to suitable sessions. Microsoft's experimentation guidance emphasizes a specific hypothesis, success and guardrail measures, and an appropriate unit of comparison.

Capacity complicates this example. A place booked by one customer is no longer available to another. Ask the analyst to design a comparison that accounts for shared class capacity; two groups competing for the same small pool of seats may affect each other's outcomes. A simple before-and-after improvement, especially after a schedule change, does not isolate the effect of the new view.

For people who cannot obtain a suitable session, clearer information may reduce enrollment while preventing a poor purchase. For people who can, the next evidence is whether they book and attend. Evaluate those outcomes separately before deciding whether to retain the change or investigate another cause.

Explore possible changes with Sapien

At Sapien, we build synthetic populations grounded in purchase, behavioral and research data. Your customer data and past research can add context. We can compare modeled reactions to alternative messages and offers, then explore the reasons different audience groups give for their responses.

For a journey investigation, that could mean comparing a membership with open booking against an offer with a defined weekly place, using complete and operationally feasible descriptions. The comparison can help prioritize explanations or alternatives for further work.

A modeled journey is a hypothesis about how someone might respond. It cannot establish that Maya visited the studio, that a seat was available or that a customer attended. Our synthetic versus human respondents guide explains the handoff between modeled reactions, first-hand accounts and behavioral evidence.

Bring us the purchase that stalls, the evidence you already have and the change your team needs to choose. We can explore audience responses and the alternatives worth investigating alongside the human and behavioral evidence the decision requires.

Book a demo.