An empathy map is a template for organising what you know about one user into four quadrants: what they say, think, do and feel. It was created by Dave Gray of XPLANE and popularised through the 2010 book Gamestorming. Its purpose is to surface the contradictions between those quadrants, particularly between what a person says and what they actually do, because that gap is where the useful findings are.
After you have done interviews or observation and need to make sense of the notes with other people. It is a synthesis tool, not a research tool: an empathy map filled in from a meeting rather than from evidence is a record of what the team already assumed.
What an empathy map is
An empathy map is a template for organising what you know about one user into four quadrants: what they say, what they think, what they do and what they feel. It was created by Dave Gray of the consultancy XPLANE and popularised through the 2010 facilitation book Gamestorming.
Its purpose is not to summarise a user. It is to put four kinds of evidence side by side so that the contradictions become visible, and the most valuable contradiction is between what somebody says and what they actually do.
A user who tells you their current tool is fine, and who has quietly built three spreadsheets around it, has told you something that neither quadrant reveals on its own.
- One person, one situation Not a segment
A map of "our users" averages away every contradiction, and the contradictions are the output.
- Four quadrants, two kinds of evidence Observed vs inferred
Says and Does are evidence. Thinks and Feels are your interpretation, and honest maps label them so.
- The say-do gap is the finding What you came for
If nothing in the map contradicts anything else, either the research was thin or you filled it in from memory.
- It is a staging post, not an output Between research rounds
A contradiction is a lead. Turn it into the next interview question rather than into a conclusion.
Why it matters
Interview notes are unusable in raw form. Twenty pages of transcript contain the finding somewhere, and nobody on the team will read them twice. The map exists to compress that into something a group can look at together and argue about.
The failure it prevents is subtle: a team that never lays the quadrants side by side will reliably build for what users said, because that is the part that gets written down and repeated. What they did is harder to record and much more predictive.
The give-away phrase is “users told us”. It is almost never followed by a description of anything a user did.
The four quadrants
Verbatim quotes from research. Their words, not your summary.
"Honestly the process is basically fine, we just get a lot of feedback."
Behaviour you observed, including the workarounds nobody mentions.
Keeps a private spreadsheet of themes. Missed the roadmap deadline twice this quarter.
What occupies their mind that they would not say out loud. An inference.
Worries that engineering thinks their prioritisation is arbitrary.
Emotional state, as an adjective plus context. Also an inference.
Defensive about the current process, because they designed it.
Read Says against Does. Here the user reports the process is 'basically fine' while maintaining a private spreadsheet and missing the deadline twice. That gap is the finding, and neither quadrant states it alone.
When to run it
- You have interview notes and nobody has read them twice.
- The team disagrees about what a user actually needs and cannot resolve it.
- You want to hand research to people who were not in the room.
- You suspect what users say and what they do have diverged.
- You are preparing a second round of interviews and need better questions.
- You have not done any research yet. Use The Mom Test →
- You need a profile the whole company can act on. Use Buyer Persona Canvas →
- You want to know what they are hiring a product to do. Use Jobs to be Done →
- You need structure for observing in the field. Use AEIOU →
Empathy mapping in practice: Google Glass
An empathy map has quadrants for what the user sees and feels. This is what happens when you fill them in for the user and nobody else.
Google Glass · 2013 to 2015
Launched to the people most excited about it, who turned out to be the wrong first users.
The Explorer programme put Glass on developers and enthusiasts at $1,500. They were the tier most willing to buy, which made them look like the ideal beachhead and made them precisely the wrong one: their enthusiasm was for the device, and the public reaction was to the camera on the face of the person wearing it.
Within two years the term "Glasshole" was in general use, venues were banning it, and the consumer programme was withdrawn in January 2015.
Glass then found a real market in enterprise, on factory floors and in warehouses, where a hands-free display solves a stated problem and nobody is offended by a colleague wearing one.
- Explorer price
- $1,500
- Consumer programme withdrawn
- January 2015
- Where it actually worked
- enterprise and industrial
What it shows: A launch segment is not the group most excited by the technology. It is the group whose problem the technology solves, in a setting where using it carries no social cost. Those are frequently different groups.
Empathy maps vs the alternatives
What do we know about this person, and where does it contradict itself?
Gives you: Four quadrants and the gaps between them
How do I structure what I record in the field?
Gives you: Observation notes in five categories. Feeds a map
What does the person who signs need to believe?
Gives you: Five rings of buying insight, aimed at commercial decisions
What happens over time, across touchpoints?
Gives you: A sequence. Empathy maps are a snapshot by comparison
When it won’t help you
- A map built from memory records assumptions, not users
A team can fill all four quadrants in twenty minutes without any research and produce something that looks exactly like a finding. It cannot contradict what the team believed, because it came from what the team believed.
Instead: Mark every entry with its source. Anything without one is a hypothesis, and labelling it so is what keeps the map honest.
- Two of the four quadrants are interpretation
Thinks and Feels are your reading of someone else's inner state. Presented in the same box as verbatim quotes, they acquire an authority they have not earned.
Instead: Visually separate evidence from inference, and treat the inferred quadrants as questions for the next round.
- It is a snapshot, and behaviour is a sequence
The map holds one moment. It cannot show that the user's state changes across a workflow, or that the frustration only appears at the third step.
Instead: Use a journey map when the question is about what happens over time.
- It does not tell you what to build
A well-made map produces understanding and no decision. The gap between "we now understand this person" and "therefore we should build X" is not something the format helps you cross.
Instead: Pair it with a prioritisation instrument. Understanding is upstream of a roadmap, not a substitute for one.
Further reading
- Dave Gray, Sunni Brown & James Macanufo, Gamestorming (2010). Where the four-quadrant version was popularised.
- Dave Gray’s 2017 Empathy Map Canvas, which adds goals, pains and gains around the original quadrants.
- The Mom Test. How to gather the material a map synthesises.
- AEIOU. A structure for capturing observation in the field.
- Buyer Persona Canvas. Where the synthesis turns into commercial decisions.
- Customer Journey Mapping. When the question is about sequence rather than snapshot.
How to apply Empathy Maps
- 1
Pick one specific user, not a segment
One person, or one tightly-defined role in one situation. A map covering 'our users' averages away every contradiction, and the contradictions are the output. Write the name and the situation at the top.
- 2
Fill Says with verbatim quotes only
Their words, not your summary. The moment you paraphrase, you have replaced what they said with what you concluded, and the quadrant can no longer disagree with the others.
- 3
Fill Does from observation, not inference
What you watched them actually do, including the workarounds. If you did not observe it, mark it as an assumption rather than writing it as fact. This is the quadrant that makes the map worth building.
- 4
Fill Thinks and Feels, and mark them as inference
These two are interpretations, and honest teams label them as such. What occupies their mind that they would not say out loud, and what emotional state are they in? Both are your reading, and both should be testable.
- 5
Hunt for the contradictions
Read Says against Does. Someone who says the current tool is fine and has built three spreadsheets around it has told you something neither quadrant said alone. If nothing contradicts, either the research was thin or the map was filled in from assumption.
- 6
Turn the gap into a question, not a conclusion
A contradiction is a lead, not a finding. Write the next interview question it implies and go back out. The map is a staging post between rounds of research, not a substitute for the second round.
Common mistakes
- **Filling it in as a team, from memory.** The most common failure by a wide margin. A map built in a meeting records what the team already believed and cannot contradict it, which is exactly what it exists to do.
- **Mapping a segment rather than a person.** 'Our SMB users' averages away every contradiction. One person, one situation, or the exercise has nothing to find.
- **Paraphrasing in the Says quadrant.** Your summary is your conclusion wearing their voice. Verbatim or nothing.
- **Treating Thinks and Feels as observed.** They are inferences and should be labelled as such, otherwise the map presents your interpretation with the same authority as your evidence.
- **Stopping at the map.** It is a synthesis step between research rounds. A completed map that nobody turned into a follow-up question was a formatting exercise.
- **Building one per persona and filing them.** Maps go stale like any research artefact, and an undated one gets cited years later as though it were current.
How ShipFit operationalizes this
ShipFit runs Ethnographic Empathy Maps in Stage 2 (Who Pays?), where the buyer profile carries a verbatim statement of what is on the buyer's mind alongside their tools, budget and buying patterns. That is the Says quadrant with sourcing attached. Stage 3 (What Hurts?) then scores problems by frequency and intensity, which is where the Does and Feels material becomes a ranking rather than a description.
ShipFit runs 55 frameworks across 9 decision stages
Empathy Maps is one tool in a bigger toolkit. The full library covers market sizing, buyer discovery, MVP scoping, pricing, and launch.
The Mom Test
Q3Rob Fitzpatrick
Validation question methodology, real interviews, not theater
Jobs-to-be-Done
Q2-Q4Clayton Christensen
Functional, social, and emotional jobs your product fulfills
7 Powers
Q4Hamilton Helmer
Strategic moats: Scale, Network, Counter-positioning, Switching, Brand, Cornered Resource, Process
Van Westendorp PSM
Q6Feature-weighted price sensitivity analysis without guessing
Blue Ocean Strategy
Q4Kim & Mauborgne
ERRC framework: Eliminate, Reduce, Raise, Create
Fake Door Testing
Q7Pre-build behavioral validation with landing pages and apology modals
+ 49 more: TAM/SAM/SOM Analysis, Porter's Five Forces, Market Timing Analysis, Unit Economics (LTV/CAC)...
Frequently asked questions
What are the four quadrants of an empathy map?
Who created the empathy map?
What is the point of an empathy map?
Should I make one empathy map per persona?
Can I fill in an empathy map without doing research?
Keep exploring
The 9-step playbook from market verdict to ship-ready spec.
Four survey questions, four cumulative curves, four intersections. How to run the Van Westendorp price sensitivity meter, plot it, and read the price range.
The JTBD framework in plain terms: the four forces of progress, the switch timeline, and how the Christensen and Ulwick schools actually differ.
Most founder market research is a TAM slide that nobody believes. The numbers that actually matter are smaller, harder to defend, and tell you whether the market exists for the ten-customer version of your business.
Most founders confuse idea validation with idea-receiving-encouragement. The two have nothing in common. Here's what real validation looks like, and the four methods that actually produce it.
5 questions. 0-100 score. One brutal verdict on whether to build.
The smallest version of a product that lets you test a single, falsifiable hypothesis about a buyer's behavior. Coined by Frank Robinson in 2001; popularized by Eric Ries in The Lean Startup (2011). Often not a working product at all (a landing page, a Wizard of Oz prototype, a manual concierge service). The defining feature: it can fail. If your MVP can't fail, it's not an MVP, it's just a small launch.
Validation for solo founders with no cofounder to push back. ShipFit forces 9 decisions and argues with your idea in 2 minutes on real data. Start free.
IdeaProof scores your idea in 120 seconds and bundles brand assets, AI ads and landing pages. ShipFit forces 9 sequential decisions with live G2 / Trustpilot / Reddit signal, Van Westendorp pricing and exports to 7 coding tools. A score is an opinion. A playbook is a plan.
Ready to make your next product a success?
9 decisions between your idea and a product worth building.