Chapter 12 · Part 4: Customers

Designing with Empathy

“Good design is as little design as possible.”

— Dieter Rams

4:30 in the afternoon

The ownership queue tested beautifully. In usability sessions held in a quiet room at Brindlepath’s office, navigators found their patients in seconds, confirmed ownership in two clicks and rated the experience highly. Three weeks after launch at the first pilot site, usage had fallen by a third.

Lena Park went to find out why. She spent two days sitting beside navigators at the customer, from the start of their shift to the end. By mid-afternoon she understood. The navigators worked in an open-plan office with phones ringing constantly. They used aging computers with small screens, often with a headset on and a patient on the line. The queue’s confirm button was small and sat next to the reassign button; twice that afternoon, Lena watched a navigator reassign a patient by mistake and spend five minutes undoing it. Worse, the queue refreshed when new discharges arrived, so the list jumped under the navigator’s cursor in the middle of a call.

None of that showed up in the usability lab, because the lab was not the navigator’s world. The product worked. It did not work there.

The fixes were modest: a larger, separated confirm button, an undo for reassignment, and a queue that announced new arrivals instead of reshuffling. Usage recovered within two weeks and kept rising. The lesson cost Brindlepath a month. It is cheaper to learn it the way Lena eventually did: by going to where the work happens.

Empathy is a method, not a feeling

Empathy in product work is not about being kind or caring, though both help. It is a disciplined practice of understanding the people you build for well enough to anticipate what they need, often before they can say it. It has three parts:

  1. Context: where, when and under what conditions people use the product.
  2. Goals and emotions: what they are trying to achieve, what they fear, and what success feels like to them.
  3. Constraints: what limits them: time, training, technology, rules, other people.

In healthcare, the stakes of getting this wrong are high, and the people you design for are often exhausted. Clinicians already spend a large share of their day on screens. In a widely cited time-and-motion study of 57 physicians in four specialties, Christine Sinsky and colleagues found that only 27 percent of the office day went to direct clinical face time with patients, while 49 percent went to the electronic health record and desk work: nearly two hours of screen work for every hour with patients.1

Every product sold into that environment either adds to that burden or relieves it. Empathy is how you know which.

Methods for understanding people

Contextual inquiry. Watch people do their real work in their real environment, and ask them to explain as they go. This is the single most powerful empathy method, and the one most often skipped because it takes time and travel. Lena’s two days beside navigators found more than a month of lab testing.

Shadowing. Follow one person through a whole shift or day. You see the transitions, interruptions and workarounds that no interview reveals.

Journey mapping. Map the steps a person goes through to achieve a goal, including what they do, think and feel at each step, and where your product touches the journey. Journey maps make pain points visible to the whole team, including people who never meet customers.

Diary studies. Ask participants to record their experiences over days or weeks. Useful for understanding behavior that unfolds over time, such as a patient’s recovery after discharge.

Personas, used carefully. A persona is a short description of a type of user, based on research. Personas help teams remember who they are building for. They become harmful when they are invented, or when they replace contact with real people. A persona should be a summary of evidence, revisited when the evidence changes.

Designing for the real environment

Lena’s discovery generalizes into a few principles for products used under pressure.

Design for interruption. People will be pulled away mid-task. Save state automatically, make it obvious where they left off, and never lose work.

Prevent errors, and make them easy to undo. Separate destructive actions from common ones, confirm only when an action is truly hard to reverse, and offer undo everywhere else. Confirmation dialogs that appear constantly train people to click through them.

Keep the screen still. Do not move things under someone’s cursor. Announce changes; let the user choose when to see them.

Minimize clicks and typing for frequent tasks. A task done forty times a day deserves to be fast. Use keyboard shortcuts, sensible defaults and bulk actions.

Use plain language. Write for someone who is tired and distracted. Labels should say what happens, not what the system calls it.

Design for the hardware and the setting. Old screens, shared computers, gloves, glare, noise. Test on the actual devices in the actual place.

Design for everyone. The accessibility practices in Chapter 11 make products better for people with disabilities and for everyone working under strain.

The users who never log in

In healthcare, the person who benefits most from your product often never sees it. Brindlepath’s real beneficiary was the patient who went home from the hospital and needed someone to notice if things went wrong. That patient never logged in, but the product’s design affected them directly: how quickly a navigator called, whether the navigator had the right information, whether the follow-up visit happened.

Designing with empathy for the beneficiary means asking, at every decision, how it changes the experience of the person at the end of the chain. Brindlepath’s team began each design review with a short story of a real (anonymized) patient whose care depended on the workflow in question. It changed conversations. “This adds a step for navigators” became “this adds a step, and here is what it means for Mrs. Alvarez’s follow-up call.”

The same principle applies in other industries: the student behind the teacher’s dashboard, the customer behind the support agent’s screen, the citizen behind the government caseworker’s system.

Balancing empathy and innovation

Empathy keeps products grounded in real needs. Taken too literally, it can also keep them timid. Users describe their problems in terms of what they know, and their suggestions tend to be small improvements to today’s workflow. Breakthrough products often change the workflow itself.

The answer is not to ignore users, but to separate their problems, which you should trust deeply, from their proposed solutions, which you should treat as one input among many. Brindlepath’s navigators asked for a better spreadsheet. What they needed was for ownership to be assigned before they arrived in the morning, which no spreadsheet could do. The insight came from empathy; the solution came from the team’s knowledge of what technology could do.

A few practices help balance the two:

  • Test bold ideas with low-fidelity prototypes, so users react to the concept rather than the polish.
  • Introduce change gradually where workflows are safety-critical, with the option to use the old way during transition.
  • Measure whether the innovation actually relieves the burden, using time-on-task, error rates and the users’ own sense of workload.
  • Involve users as partners, not just test subjects. Brindlepath formed a small advisory group of navigators from three customers who saw designs early and helped shape them.

Continuous improvement for user-centricity

Empathy is not a phase. People’s work changes, products change, and the gap between them opens again. Keep it closed with a few lasting habits:

  • Regular field visits for every product trio, at least once a quarter.
  • A steady stream of feedback from support tickets, in-product feedback and customer success, reviewed by the product team weekly.
  • Usability measures for key tasks, tracked over time: completion rates, time on task, error rates.
  • Satisfaction measures used with care. Net Promoter Score and similar surveys can show trends, but a score without the stories behind it says little about what to fix.

Brindlepath in practice: the navigator advisory group

Lena’s discovery in the opening scene led to a lasting practice. Brindlepath formed a navigator advisory group: eight navigators and two navigator managers from four customers, who met with the product team for an hour every month and were paid a modest stipend, with their employers’ agreement, for their time.

The group saw designs early, usually as sketches or rough prototypes, before the team had invested much in any direction. They tested new workflows in their own environments for a week before wider release. And they brought problems the product team had not heard about: the navigator who covered a rural clinic with an unreliable connection, the manager who needed to rebalance workloads when someone called in sick, the new hire who could not tell which patients were most urgent on her first day.

Two rules kept the group useful. First, the product team presented problems and options, not finished solutions, and asked the group to argue about them. Second, the team closed the loop: every month began with a short review of what had changed because of the group’s input in previous months. Members said this was the reason they kept coming. Seeing their suggestions in the product made them advocates inside their own organizations, which customer success noticed in renewal conversations.

The group did not replace field visits or broader research. Ten people cannot represent every navigator. But it gave the product team a standing relationship with real users, and it caught several problems, including a confusing change to how risk levels were displayed, before they reached hundreds of navigators.

Accessibility in a real workflow

When Brindlepath ran its first full accessibility review of the ownership queue, it found problems no usability session had surfaced. The risk level was shown only by color, so a navigator with color blindness could not tell high from medium risk. The queue could not be used with a keyboard alone. A screen reader announced each patient row as “button, button, button.”

None of these was hard to fix. Risk levels gained text labels and icons. Every action gained a keyboard shortcut and a visible focus state. Rows were labelled with the patient’s name, risk level and discharge time. The fixes took one engineer two weeks, and the team added the checks to its definition of done so new screens would not repeat the mistakes. The keyboard shortcuts became one of the most popular features in the product, among navigators with and without disabilities.

Common mistakes

Testing only in the lab. Quiet rooms, fast computers and focused participants hide the interruptions, noise and fatigue of real work.

Inventing personas. Personas made up in a workshop reflect the team’s assumptions. Build them from research and revisit them when the evidence changes.

Designing only for the logged-in user. Patients, students, citizens and customers who never see the product are often the people it most affects.

Taking users’ solutions literally. Users are experts in their problems. Their proposed solutions are one input, often a modest improvement on today’s workaround.

Adding confirmation dialogs everywhere. Constant confirmations train people to click through them. Prevent errors by design and offer undo instead.

Equating empathy with satisfaction scores. A score shows a trend; it rarely tells you what to fix. Stories and observation do.

Try this

Arrange for every member of your product trio to spend half a day with a user in their real environment within the next month, watching rather than demonstrating. Each person should come back with three observations: one workaround, one moment of frustration and one moment where the product helped. Combine them into a simple journey map like Figure 12.2. Then pick the most painful moment on the map and make it the subject of your next round of discovery.

How AI changes this

AI can reduce burden, or add to it. The most promising AI features in healthcare, such as ambient documentation that drafts clinical notes from a conversation, aim directly at the screen time in Figure 12.1. But an AI feature that produces output clinicians must carefully check adds burden of a different kind. Design AI features around the reviewer’s time: make outputs easy to verify, highlight uncertainty, and make correction fast.

AI can scale empathy research. Assistants can transcribe and tag hours of field recordings, cluster support tickets by emotion and theme, and draft journey maps from research notes. That leaves more time for the field visits themselves, which no tool replaces.

Synthetic users are not users. As Chapter 3 warned, AI-generated personas can rehearse an interview guide, but they cannot tell you what it is like to be a navigator at 4:30 in the afternoon. Lena’s lesson does not change.

AI interfaces need empathy too. Conversational and agent-based features raise new design questions: how much the AI should do on its own, how to show what it did and why, how to let people stay in control, and how to avoid sounding more certain than it is. Chapter 17 covers these in depth.

Chapter checklist