Learn how to improve customer experience in your company: pick one journey, find why it stalls, test a fix and measure it, with one imagined request.
To improve customer experience, pick one journey, find out what stops customers finishing it, and test a change you can measure before you scale it. The cause may be a vague message, information that never reaches the next team, or a step that sends customers back to a point they've passed. Ask where the customer is doing work the company could do for them.
Say a customer submits a request online and gets a confirmation. Two days later they call about the appointment, explain everything again and get passed to another team. Your reports show a working website and an answered call. Their first question is still open: when will this get done? We'll follow this imagined example throughout.
Where should a company start improving customer experience?
Start with one journey that has a clear beginning and end, and gather evidence of what actually happened before anyone proposes a fix.
What does customer experience cover in your company?
Customer experience starts before anyone contacts support and runs through every channel the company offers. It's wider than customer service, as What Is Customer Experience? explains. Kim Flaherty of Nielsen Norman Group described a customer journey in 2021 as “the end-to-end process that a customer goes through to complete a task over time”, and each interaction in it may belong to a different part of the company. So this work has to cross the boundaries of any one product or department.
In our example, the agent may earn a good rating for being clear and helpful while the customer stays unhappy about the wait. That's why you first define a successful end to the journey. “Request submitted” is a fair finish line for a form, not for a customer who wants the service. Write the outcome in their words: “I knew my appointment, the service got done, and I didn't have to keep chasing.”
Keep the problem apart from its proposed solution. “We need a new app” is an early technical decision. “Customers can't tell where their request stands” is a problem you can solve in more than one way, and fixing an existing procedure may be enough.
Start with one journey your team can follow
Customer experience optimisation is easier to run as one journey than as a programme. “Improve our relationship with customers” is too broad for daily work. “From a maintenance request to confirmed completion” is a scope you can review with customers and staff, across the website, the phone and a site visit.
To choose, go through complaints and operational data for a journey where people keep returning to support or wait without explanation. Weigh the harm to the customer against how easily your team can reach the evidence. Volume alone won't tell you enough, since some customers may leave without complaining.
Then write a short starting sheet:
The customer: who goes through this journey, and when?
Start and end: what event proves it's complete?
The observed problem: what's the evidence, and what's still an assumption?
The owner: who brings people together and follows the decision through?
The baseline: what will you compare against later?
If the team disagrees on what “complete” means, settle that first. Otherwise one team can celebrate fewer calls while customers can't reach help at all.
Gather evidence that explains what happened
Take a sample of real cases, stripped of personal data you don't need. Follow each from the request to the last contact, and record the waits, the messages and the transfers between teams.
Then talk to customers who went through the journey. Ask about the last time they used the service: what they expected after submitting, and what made them call. A general question about satisfaction won't rebuild what happened.
Give the employee who takes the request room to explain their work too. They may have to ask again because the information doesn't reach their screen, or because they can't confirm the appointment. Knowing that limit makes for a more realistic recommendation than “be faster”.
Keep apart what you saw in the record, what the participant said and what the team concluded. Repeated calls are a fact you can verify. Reading them as lost trust needs more evidence.
Collect cases that finished as well as cases that stalled: the comparison can show a condition the complaints didn't, such as the request type or the channel. Don't present a handful of interviews as percentages of your customers. If the stall starts before your scope, move the scope rather than forcing the findings to fit.
What the record shows is one thing; your reading of it needs its own evidence.
The journey ends when the customer gets what they came for, not when the form is sent.
How do you improve customer experience once you have the evidence?
Connect what the customer sees to what happens inside the company, choose one change that addresses the cause, and test it on a small group before you extend it.
Find where the journey breaks between teams
Map the customer's steps in a customer journey map. When you need to see what happens inside the company at each step, use a service blueprint. Sarah Gibbons of Nielsen Norman Group defined it in 2017 as “a diagram that visualizes the relationships” between people, physical or digital evidence, and processes, all tied to the touchpoints of one customer journey.
In our imagined example, the customer sees “We've received your request”. Behind that line, the request has to reach the right team, the data has to be checked, the appointment booked and a clear message sent. Place each internal step under the event the customer sees, and note where responsibility changes hands.
You may find the message goes out before anyone owns the request. Rewriting it won't help then: you need to know who reviews it and what happens if nobody picks it up in time. Gibbons notes that poor experiences are “often due to an internal organizational shortcoming”, so the cause can sit outside the screen altogether, as we explain in Systems Thinking in Design.
Gibbons adds that when information gathered early is reused later, the customer isn't asked the same question twice.
Don't fill a blank on the map with a tidy guess. If you don't know how the status gets updated, write that operations needs to confirm it. A visible gap is more useful than a neat step that doesn't really happen.
The diagram below lays out the same example in three stages, and each stage shows what the customer sees, what happens inside the company and who owns it. At the review stage, for instance, the customer sees the request's status and next step while the service coordinator checks the data and sets the appointment. If nobody was assigned the request when it came in, the customer is left with “We've received your request” and no next step, and that's when the calls about the appointment start.
Choose a change that addresses the cause
Initiatives to improve customer experience will now pile up: a better confirmation message, a visible status, fewer transfers, two systems connected. Rank them by the problem they solve, not by how well they present. Ask how much harm the problem does, how strong the evidence is and what a test would take.
Say the case review shows customers call because the confirmation doesn't state the next step, while the appointment is already in the system. Try a clearer message; the wording is part of the experience, as When Messages Become Part of the Customer Experience explains. If no appointment has been set yet, showing one isn't an option.
Write the change as a hypothesis: “If we show the confirmed appointment and how to change it, we expect fewer calls asking about it.” Then state what would disprove it: calls might stay high because the link doesn't work on customers' devices.
Don't add a promise operations can't keep, and check what the customer sees if the request is rejected, data is missing or no appointment is available. Decide who changes the message, who approves the published timeframe and who fixes a failed send. A name next to each decision stops the meeting ending with a list of improvements nobody can carry out.
Test before you generalise
Set up a limited test: a prototype of a new message, or a new procedure on a small group of cases. Decide who watches it and when to stop if it causes confusion or extra errors, and record the version you used.
Decide what you'll test before you show a result, and connect research to design testing, as the MOLIM case study sets out. After launch, check the operational data to see whether the problem that prompted the change actually fell. Each stage has its own question and its own evidence.
What should you measure to know the experience improved?
Choose your measures after you've defined the problem. If customers call to find out their appointment, start with the share of requests followed by a call for that reason and the time it takes to reach a confirmed appointment. Then add one question on the experience, rather than a long survey unrelated to the decision you're testing.
What each metric tells you
Customer Satisfaction Score (CSAT): in IBM's 2024 explainer, satisfaction with a specific product, service or interaction.
Customer Effort Score (CES): introduced by Matthew Dixon, Karen Freeman and Nick Toman in Harvard Business Review in 2010, it measures how much effort a customer has to put in to deal with a company or its products. IBM lists agent touches among the measures that help predict it: the more agents a customer passes through, the more likely they are to rate the experience as high effort, as in our example.
Net Promoter Score (NPS): how likely customers are to recommend you, on a zero-to-ten scale. In Bain & Company's method, subtract the percentage who answer 0 to 6 from the percentage who answer 9 or 10. It doesn't tell you how many customers completed their request.
Keep the question wording, the scale direction and the send time fixed so comparisons make sense.
A scorecard for the appointment example
Outcome: the share of requests with a confirmed appointment within the published time.
Effort: repeat calls about the appointment, against eligible requests.
Customer view: one question on how easy it was to find the appointment, after the confirmation.
Side effect to watch: requests closed before the problem was solved, or wrong appointments corrected.
These are proposed measures for the example, not project results. Define the denominator of every percentage, compare periods or groups that are as similar as you can get, and don't credit every improvement to the change. A lower average handling time can hide a few cases that waited far too long, so choose a view that shows them too.
Decide, then hand the change over
Review the evidence and decide: extend, adjust or stop. Watch for work that has simply moved to another employee, and record why rather than settling for “the test worked”.
Before you extend it, test the handover with someone who had no part in designing it. Give them the procedure and the messages, and ask them to follow one normal case and one exception. If they need a long verbal explanation, add what's missing to the work instructions. Tell the support agent who reviews a mismatch between the message and the system, and what holding reply to give. Success shouldn't depend on the designer standing next to every employee.
How does improvement continue after the project ends?
A journey needs an owner who reviews its evidence and follows its decisions. In the same Nielsen Norman Group article, Flaherty describes journey management as “the ongoing practice of researching, measuring, optimizing, and orchestrating a customer journey”, and says a journey manager should own “just a single customer journey”.
Set a short, regular review of a recent real case, the measures and the open decisions. Keep a simple log of each change and what the team learned, so the same debate doesn't reopen without the evidence.
If you're asking how to improve customer experience in your company right now, start with one request that stalled recently. Follow it from the customer's side, bring together the people who handled it, and pick one point worth changing. Then agree what evidence would make the case for extending it.
If you'd like support researching journeys and connecting them to operations and measurement, see AZM X's customer experience consulting. Bring the journey you want to review and whatever evidence you have to a conversation with our team.
