What users say and what they actually experience are rarely the same thing
A collaborative synthesis tool that organises research observations into four quadrants — Think, Feel, Say, Do — to surface what's really going on for a user, and to make that understanding shared across a team.
What is an Empathy Map?
An empathy map is a research synthesis tool that organises what you know about a specific user into four quadrants: what they Think, what they Feel, what they Say, and what they Do. The goal isn't to document everything — it's to surface the gaps between stated behaviour and actual behaviour, and to translate raw research into something a whole team can reason from.
Dave Gray at XPLANE developed the original empathy map format and has refined it several times since. The version most teams now use adds Pains and Gains below the four quadrants — influenced by the Jobs To Be Done tradition — to capture what frustrates and what motivates the user in their broader context.
The reason empathy maps exist is a communication problem. When one researcher spends two weeks in the field, the insight lives in their head. The map turns that understanding into a legible artifact that product managers, engineers, and stakeholders who weren't in the sessions can actually use.
Empathy Map vs Persona: Not the Same Thing
This confusion is common and it matters. A Persona describes who your user is — role, goals, background, and typical behaviours drawn from patterns across many users. A persona is a composite character.
An empathy map captures what a specific user was experiencing in a specific context at a specific point in time. It's raw rather than synthesised. You can build an empathy map from a single interview session. You shouldn't build a persona from one.
The two complement each other. Empathy maps feed personas — you map individual sessions, then look across multiple maps to find patterns that become persona characteristics. But if your team uses the terms interchangeably, you've merged two distinct tools into one vague concept and lost the precision of both.
The Four Quadrants (and the Two Most Teams Rush Through)
The quadrants sound obvious until you try to fill them honestly:
- Think: The user's inner monologue. What beliefs, assumptions, and worries are they carrying into the experience? What are they not saying out loud?
- Feel: What emotions surface during the task? Frustration, confusion, confidence, relief? This quadrant is where research gets uncomfortable — and where insight usually lives.
- Say: Direct quotes, not paraphrases. "I just click around until something works" is more useful than "user lacks confidence in navigation."
- Do: Observable actions. What do they actually click? What do they ignore? What workarounds do they create?
Teams fill Say and Do quickly because those are easy to observe. Think and Feel require interpretation — and that interpretation is where the real synthesis happens. Rushing past those two gives you a behavioural log, not an empathy map.
When to Build One — and When Not To
Empathy maps are most useful right after primary research — Contextual Inquiry sessions, discovery interviews, or diary studies — when you have dense qualitative data that needs to become a shared team artifact. Build them while sessions are fresh, ideally the same day. The texture of what a user said and felt fades fast.
They're not useful when:
- Built from assumptions rather than research. A team filling in a map based on what they think users experience is just a structured guess. That's not synthesis.
- Built three weeks after the sessions. A map created from memory is a memory of a memory.
- Treated as a deliverable rather than a tool. The map isn't the point — the conversation it triggers is.
One signal you're using it wrong: if your team builds an empathy map and nothing about the design changes as a result, it wasn't built with enough specificity.
What Good Looks Like
A good empathy map:
- Uses direct quotes in the Say quadrant, not cleaned-up summaries
- Covers a real user or a clearly defined research segment — not "our target customer"
- Shows tensions between quadrants: what someone says doesn't match what they do; what they feel contradicts what they think they should feel
- Gets challenged and updated in a team session, not built in isolation by one researcher
The most useful output of a good empathy mapping session is a "we assumed X but the research shows Y" moment. If those moments don't happen, the team is confirming what they already believed — which is a waste of research.