011 min
Why teams draw empathy maps
Picture a product team that has interviewed six people about an app for sharing household bills. This is an illustrative case, not a real product. The product manager remembers that people asked for reminders. The designer remembers that people worried about being blamed for a late payment. The engineer remembers a complaint about notifications.
All three are right, and the team still cannot decide what to build, because each person holds a different part of the same research. Nobody can say which part matters most, or whether a fourth part is missing.
The team's answer is an empathy map, drawn on one page for one type of user. Each person adds what they learned to the same page, so the team argues about evidence instead of memories. The page also shows what nobody has learned yet, because an area with no notes is an area the interviews never reached.
022 min
How an empathy map is laid out
An empathy map has a person drawn in the middle and four areas around the person: Says, Thinks, Does and Feels. The layout shows one mindset at one point in time. It is not a sequence of steps.
| Area | What goes in it | Where the evidence comes from |
|---|---|---|
| Says | Quotes, as close to the user's own words as possible | Interview and session transcripts |
| Thinks | What the user thought, including what they did not say aloud | Read from tone, hesitation and the questions they asked |
| Does | Actions the team saw the user take | Observation, recordings and usage records |
| Feels | Emotions, each with the reason for it | Read from expression and context |
Two of these areas are observed and two are read from clues. A team hears what a user says and sees what a user does. Nobody can watch a thought or a feeling, so the team reads them from hesitation, tone and choices, and writes the clue next to the reading. The Thinks area has a special job. It holds what the user thought but did not say, perhaps because they were unsure, polite or self-conscious.
Specific entries are worth more than general ones. A quote that only says the user wants something reliable could come from a user of almost any product. A quote that names the moment things went wrong shows the team where to look. The same goes for feelings. The word worried gives the team nothing to act on, while worried that the screen does not show whether the other person has paid does.
Many teams add areas. See and Hear record what the user is exposed to, such as other people's remarks or what they read. Pains and Gains record what the user fears and hopes for. Goals, a common fifth area, ties the other four to what the user is trying to achieve. A map can describe one real person or a segment, which is a group of users who behave alike. Teams usually start with one person and build a segment map by combining several individual maps.
032 min
Why empathy map areas disagree
The Says and Does areas of an empathy map often disagree, and the cause is how people report on themselves. The map therefore fills the two areas from different evidence. Says comes from what people told the team. Does should come from what the team saw, such as a recording or a usage record, because people also report their own actions from memory. When the two disagree, the team has found something to investigate. The evidence behind this is below, for readers who want it.
Three ways self-reports are wrong
Jakob Nielsen wrote in 2001 that people's reports of their own past behaviour are typically wrong in three ways. The first is social: people adjust their answers toward what they think the listener wants to hear. The second is memory: people report what they remember doing, and memory loses the small details. The third is explanation: people give reasons for their behaviour afterwards, and the reasons sound reasonable even when they were not the cause.
A survey answer that sounded like demand
Nielsen's example was a survey in which 50% of respondents said they would buy more online from shops that show 3-D product views. He read the number as a sign that 3-D sounds attractive, not as a sign that people would use it. Guesses about future behaviour are weaker still than reports of past behaviour.
Liking an idea is not using it
Drop-down menus show how this goes wrong. In 2001 Nielsen reported that users loved the idea of them, yet the menus often confused users or sent them to the wrong part of a site.
Opinions predict little of performance
Nielsen and Jonathan Levy analysed 113 comparisons of interface pairs built for the same task and found a correlation of 0.44 between measured performance and stated preference. A later study of websites by Nielsen found 0.53. At either value, what people say they like predicts only about a quarter of how well a design works.
042 min
Where empathy maps come from
Empathy maps began as a short team exercise rather than a research method. Dave Gray of the design consultancy XPLANE developed the empathy map in the 2000s as a way to turn research notes into a picture a whole team can discuss. It reached a wide audience as one of the exercises in Gamestorming, a collection of facilitation games. Alex Osterwalder's book Business Model Generation later presented it as a tool for finding insights about customers.
The first published exercise was short. Someone draws a circle on a board for one stakeholder, names it, and divides it into areas for what that person thinks, feels, says, does and hears. The page that describes it says a rough map takes about 20 minutes. It does not ask the team to collect research first. It asks them to use what they already know about the person, and then to ask others to review the map.
Why the original exercise draws a face
The page treats the face as part of the method, not as decoration. It says the small details of eyes, hair and glasses help the team imagine being that person, which is the aim of the exercise. A team that cares about who the map describes is more likely to keep that person in mind while filling it in.
From imagination to evidence
Current guidance has moved away from imagination. The Nielsen Norman Group asks that each note on a finished map can be traced to a source in the research, such as a transcript or a recording. It also asks that a note with no source be labelled an assumption.
052 min
How to build an empathy map
A team that wants a map it can trust now starts with research, and the steps below show one order for a workshop.
- Choose one person or one type of user. A map for all customers mixes people who have different goals.
- Collect qualitative research: user interviews, field studies, diary studies and open-ended survey answers. Support tickets and call logs add complaints that people raised without being asked.
- Have each team member read the research alone and write one note at a time, placing each in one area. Working alone first stops the first opinion spoken aloud from setting the group's view, an effect called anchoring.
- Put the notes on a shared board and group similar ones within each area. Name each group. Then ask which notes fit no group, which themes repeat across areas, and which areas have few notes.
- Write the date, the version and the open questions on the map, so a later reader knows how current it is.
A map can also be a research tool: at the end of an interview, hand the participant a blank template and ask them to fill in the four areas for what they just did.
When to ask users
The Says area is only as good as the moment the team asked. Nielsen advised collecting preference data only after people have used a design, because only then do they know how well it supports them. Asked before they have tried it, people comment on surface features.
A question that works before use
One question still works before people have used anything: why are they visiting today? It asks about motivation, which people can answer on arrival.
061 min
Reading a finished empathy map
Back in the bill-sharing example, the finished map for one household member shows a mismatch. In the Says area, the participant says the payment page is clear. In the Does area, the same participant takes a screenshot of the confirmation after every payment. The Thinks area records a worry that the payment did not go through. The Feels area records worry that the screen does not show whether the other person has paid.
Contradictions like this are the main use of the map, because they often point to something the user never put into words. Here, the plan for a reminder feature now looks like the wrong fix. The participants are not forgetting to pay. They doubt that the payment arrived. The team changes the question for its next interviews: what would a person need to see to believe a payment is done?
071 min
Empathy maps compared with related tools
The tool nearest to an empathy map is the persona, a profile of a type of user that includes goals and context. The two work together. The Nielsen Norman Group describes a map as one way to show what a team knows about a persona, not as a replacement for it. A map can also be a first step toward a persona, because a team can group the maps of similar users into a segment and describe it.
The other neighbours differ in shape and in method. A journey map follows a user through steps over time, so it shows how an experience changes. An empathy map does not follow time. An affinity diagram is a method for sorting many notes into themes, and a team can use one to group the notes in step four of the build process above.
| Tool | What it shows | Shape |
|---|---|---|
| Empathy map | A person's or a segment's mindset at one point in time | Four or more areas around a person |
| Persona | A described type of user, with goals and context | A profile |
| Journey map | An experience across a sequence of steps | A timeline |
| Affinity diagram | Themes found by grouping many separate notes | Groups of notes |
082 min
What research shows about empathy maps
The evidence for empathy maps is limited, and the authors of the studies below say so themselves.
A first test found no clear gain
Ezequiel Kahan and colleagues first tested empathy maps in a quasi-experiment, a study in which the researchers do not assign people to groups at random. The students were computer engineering students at the Universitat Politècnica de València, and the paper appeared in 2021. In that first test, the students perceived a slightly positive influence from empathy maps, but the results did not show a statistical benefit. The authors call the result preliminary and not conclusive.
A larger test with 74 students
A later paper by the same group reports a quasi-experiment run in 2021 with 74 students. Ezequiel Kahan and colleagues compared brainstorming with empathy maps, brainstorming with personas, and brainstorming alone. Students who used empathy maps showed a statistically significant effect on the number of stakeholders they identified, meaning the kinds of people affected by the product.
What kind of ideas changed
The kind of idea shifted too. With empathy maps, students produced more functional ideas, which are about what a product does, and fewer non-functional ideas, which are about qualities such as speed or security. The authors conclude that the evidence is not yet strong enough to prefer one technique over the other.
A use outside product work
Patrick Cairns and colleagues introduced empathy maps into the communication skills training of first-year medical students. Researchers interviewed 28 people, 16 medical students and 12 patient partners, about empathy map training at two UK medical schools. Both groups saw value in it, and students said it changed how they thought about empathy and patient-centred care.
A map does not prove empathy
The same study does not make a stronger claim. The authors call for more research before anyone can say that empathy maps raise empathy. They asked about attitudes and usefulness, not about whether students became more empathic. A map helps a team organise what it knows about a person. It is not proof that anyone understands that person better.
092 min
When empathy maps mislead a team
An empathy map looks the same whether a team built it from research or from opinion, and that is its main risk. These are the failures to look for.
- Built from memory. A team that fills in a map from memory records its own beliefs, and the team then relies on that record for later decisions. The tell is a note that nobody can trace to a source.
- Not used after the workshop. A map that is presented once and not opened again cannot correct a wrong assumption later. The tell is that nobody has opened it since the kickoff.
- A segment that is too wide. A map for all customers collects notes from people with different goals, and the notes contradict each other. The tell is a contradiction that nobody can explain.
- Out of date. Users and products change. The tell is a map with no date or version.
- Mistaken for a timeline. The map is a snapshot of one mindset, so it cannot show how needs change over time. For that, a team needs a journey map.
- Notes from a tool. A note the team cannot trace to research is an assumption, whether a colleague or a software tool produced it.
Before relying on a map, ask of each note whether it can be traced to a source. Treat the notes that cannot as questions for the next round of research.
101 min
Frequently asked questions
Can an empathy map describe someone other than a customer?
Yes. The original exercise was written for any stakeholder in a business, such as a client, a partner or a colleague, in a given situation such as a buying decision. The method stays the same.
Who should take part when a team builds an empathy map?
When the aim is shared understanding, everyone who will make decisions about the user should take part. Someone who skips the workshop has not seen the evidence and may disagree with the result later.
What is the difference between Thinks and Feels?
Thinks holds what the user believes or wonders, such as a doubt about a result. Feels holds an emotion and its cause. If an entry fits both, put it in one area and move on.
Which tools can host an empathy map?
Miro, Mural, FigJam and Lucidchart all offer empathy map templates. A wall, sticky notes and markers work as well, and a table with four columns is enough for one person working alone.
?4 questions
Questions people ask
Can an empathy map describe someone other than a customer?
Who should take part when a team builds an empathy map?
What is the difference between Thinks and Feels?
Which tools can host an empathy map?
§8 sources
Sources
Rachel Krause and Sarah Gibbons, "Empathy Mapping: The First Step in Design Thinking", Nielsen Norman Group. nngroup.com
"Empathy Map", Gamestorming, the page that describes the original exercise. gamestorming.com
"Empathy Map", Mutomorro, a practitioner summary of the technique's history and limits. mutomorro.com
Stanford d.school, Design Thinking Bootleg, 2018 deck. dschool.stanford.edu
Show all 8 sourcesShow fewer sources
Jakob Nielsen, "First Rule of Usability? Don't Listen to Users", Nielsen Norman Group, 2001. nngroup.com
Ezequiel Kahan, Emilio Insfran, Marcela Genero and Alejandro Oliveros, "Studying the Influence of Empathy Maps on Brainstorming for Requirements Elicitation: A Quasi-Experiment", ISD2021 Proceedings, 2021. aisel.aisnet.org
Ezequiel Kahan, Emilio Insfran, Marcela Genero and Alejandro Oliveros, "Could the utilization of Empathy Maps and Personas enhance the requirement elicitation process", 2024. e-publicacoes.uerj.br
Patrick Cairns, India Pinker, Andrew Ward, Evelyn Watson and Anita Laidlaw, "Empathy maps in communication skills training", The Clinical Teacher 18(2), 2021, abstract record. researchportal.lih.lu
