Empathy map: build it from interview notes, or it is fan fiction

An empathy map is a one-page picture of what a specific user says, thinks, does and feels, laid out in four quadrants around that person. Build it from real interview notes, so every sticky note points back to something a real person said or did. An empathy map made from the team's imagination is fan fiction about your customer: fun to write, and nobody should make product decisions from it.

Nielsen Norman Group defines it as "a collaborative visualization that captures what we know about a particular user or about a user segment." The key words are "what we know." The rest of this page is about keeping it that way.

1. Use the four quadrants for interview notes, and borrow Dave Gray's goal box

The version most people mean has four quadrants, with the user in the middle. NN/g describes them like this:

• Says: what the user said out loud, ideally verbatim quotes from research.
• Thinks: what goes through the user's mind during the experience, including things they may not be willing to say.
• Does: the user's behaviors. NN/g warns that what people report doing is not always what you observed them do.
• Feels: the user's emotions, each written as an adjective plus a short phrase that explains it.

NN/g says the map is deliberately not sequential. It captures a mindset, which is what separates it from a journey map (see section 5).

The method comes from Dave Gray and his design firm XPLANE. The Gamestorming page on the empathy map describes the original: draw a large head, name the person, decide what you want to understand about them (its example is "Why should I buy X?"), then divide the head into sections for what they are thinking, feeling, saying, doing and hearing. It says a rough map takes about 20 minutes if you already know the person and their situation well.

In 2017 Gray published an updated Empathy Map Canvas (last updated 16 July 2017, credited to Dave Gray, xplane.com). The Gamestorming post announcing it says many online versions had drifted from the original idea, and that the redesign drew on what Gray learned about canvas design while working with Osterwalder. The canvas has numbered sections you fill in order:

Use the four quadrants when you are making sense of interviews, because each quadrant maps cleanly onto a kind of note you took. Then add Gray's goal box at the top anyway. Writing down what the person needs to do, and how you would know they succeeded, stops the map from drifting into a general mood board. NN/g calls a fifth Goals quadrant a common adaptation, so you are in good company. The tradeoff: the Sees and Hears boxes are useful in a workshop, but interviews rarely give you much for them, so expect them to stay thin.

2. Fill it from interview notes, and label every guess as a guess

Every note on the map should trace back to a research artifact: an interview transcript, a recording, a support ticket, a field visit. NN/g puts the rule plainly: "If you can't point to the source, the note is an assumption, and it should be labeled as one."

This matters because empathy maps are easy to fill without data. Notice that Gray's own canvas asks "What can we imagine them saying?" and "What can we imagine them doing?" That is fine in a kickoff workshop where the goal is to surface what the team believes. It becomes fan fiction the moment those imagined lines are presented as findings. (The imagined customer always agrees with the roadmap. Real ones are less obliging.) The original Gamestorming instructions say it too: "Don't put your words into their mouth."

A practical way to do it:

1. Tag every note with a participant ID, such as "P2", and keep the link to the recording. How to take customer interview notes covers writing down what happened separately from what you think it means.
2. Sort notes into quadrants on your own first. NN/g recommends each person writes their notes privately before the group shares, so early opinions do not anchor the group.
3. Put exact quotes in Says and observed actions in Does. If someone told you what they do but you never saw it, it belongs in Says.
4. Write Thinks and Feels as interpretations with a source. "Feels: anxious about the spring rush (P2, said 'I lose sleep in March')" is evidence. "Feels: overwhelmed" with no tag is a guess.
5. Mark anything without a source with a clear label such as "ASSUMPTION", and add it to the list of questions for the next round of interviews.

If you are mapping before you have done any research, NN/g says to treat the whole map as hypotheses and plan studies to test them. Just do not present it to the roadmap meeting as research.

3. Start with one real person, then combine people who behave alike

NN/g's advice is to always start with one user or persona per map. A map can represent one actual person, or a segment, and aggregated maps are "usually created by combining individual empathy maps from users who exhibit similar behaviors."

So make one map per interview first. Then lay them side by side and group the people who did similar things for similar reasons, not the people who share a job title. Affinity mapping is the right tool for finding those groups, and how to analyze customer interviews explains why grouping by the reason behind an action beats grouping by the complaint.

Watch for contradictions in the combined map. NN/g says contradictions between quadrants are worth investigating, and that when they become impossible to resolve, the segment is too broad. An empty quadrant is also a finding: NN/g treats it as a signal that you need more research. The tradeoff is time: eight individual maps take longer than one workshop map, but they are the only way to know your aggregate describes anyone real.

4. Here is a made-up example: an owner of two bike shops

Here is a made-up example. A startup is building inventory software for independent bike shops. It interviewed six shop owners about how they decide what to reorder before spring. One of them, Dana (P2), runs two shops and shared her screen during the call. Her one-user map, built only from the interview, looks like this:

Goal (from Gray's canvas): reorder parts and bikes for spring without running out of popular sizes or sitting on unsold stock. Success would mean no lost sales in April and nothing left over from last year by June.

Says

• "I know where everything is. I don't need software for that." (P2, minute 4)
• "The second shop is the problem. I'm only there on Tuesdays." (P2, minute 11)

Does

• Opened a spreadsheet tab called STOCK_REAL_v2 that was last edited in January (P2, screen share, minute 15)
• Called the second shop's manager during the interview to check how many medium frames were left (P2, minute 18)

Thinks

• Suspects the second shop's counts are wrong but does not want to accuse the manager (P2, inferred from minutes 11 to 19, marked as interpretation)

Feels

• Anxious about the spring rush, because last March she ran out of medium frames for two weeks (P2, minute 21)
• ASSUMPTION: embarrassed about the outdated spreadsheet. She did not say this. Ask in the next round.

The useful part is the contradiction. Dana says she knows where everything is, but she had to phone her other shop to find out. That gap is the product opportunity, and it came from comparing two quadrants, not from anyone's imagination. (A phone call in the middle of a research interview is a gift. Most contradictions hide better.)

When the team lays Dana's map beside the other five, suppose three owners with more than one location show the same pattern and three single-shop owners do not. That is two aggregated maps, not one, because the single-shop owners have a different problem.

5. Use the map to feed personas and journey maps, not to replace them

Treat an empathy map as a working step toward other artifacts. NN/g lists treating it as a one-time deliverable among common mistakes, says aggregated empathy maps can be a step toward personas, and says just as clearly that "empathy maps are not a replacement for personas."

The difference from a journey map is time. NN/g's comparison of UX mapping methods by Sarah Gibbons describes the empathy map as focused on one type of user with no timeline, while a customer journey map follows a customer's path through a specific product or service in order, to find the moments that cause pain or delight.

A sensible order for a small team:

1. Empathy maps from each interview, then aggregated by behavior, to agree on who these people are.
2. A persona that explains why one of those groups buys. Buyer personas covers writing one that explains the purchase decision, not the job title.
3. A journey map for that persona's path through a specific decision. Customer journey mapping walks through building one from interviews.

Update the map as new interviews come in. NN/g also lists a stale map as a common mistake and suggests putting a date and version number on each one.

Your next step

Take your last three interview transcripts and make one empathy map per person, tagging every note with a participant ID and labelling anything without a source as an assumption. The empty boxes and the assumptions together are your question list for the next round of interviews. If you need more people to talk to, Instant Expert can find people matching a specific description, such as "owners of independent bike shops with two or more locations", and send them paid invitations for a short call or a written answer; you pay only if someone books or answers. For help running the research itself, see the directory pages for user research experts and UX consultants.