Empathy Map
Get Inside Your Customer's Head in Four Quadrants
The Empathy Map is the whiteboard tool everyone reaches for when they realise they've been designing for a customer they've never actually met. Four quadrants - Says, Thinks, Does, Feels - and a simple promise: stop guessing what's in someone's head and start writing it down where the whole team can argue about it.
EMPATHY MAP
“Four quadrants, one human, one situation: if you can't source a quadrant from something a real person said or did, you're guessing - and the map will show it.”
Done well, it's the fastest way to turn a pile of messy interview notes into one shared picture of a real human. Done badly, it's four sticky-note clusters of pure invention, dressed up as research, that make a room feel like it understands a customer it has never spoken to.
The Empathy Map doesn't generate insight. It organises it - and, more usefully, it shows you the gaps. An empty quadrant is the map doing its job: it's telling you which question you can't yet answer. This page walks through each quadrant, the question that unlocks it, and how to tell an Empathy Map built on evidence from one built on vibes.
What is Empathy Map?
Four quadrants around a single person and a single situation: Says (what they actually voice out loud), Thinks (what occupies their mind but they won't say), Does (observable actions and behaviours), and Feels (emotional states, with the trigger). Fill it from real research, not assumptions. An empty quadrant isn't failure - it's a research to-do. Use it to align a team on who you're designing for before you build anything.
Worked Examples
Three real brands. Different categories, different sizes. Same framework, filled in.
Monzo
UK challenger bank (founded 2015)A textbook case of designing from the Thinks and Feels quadrants. Monzo built its early product around the unspoken anxiety people feel about money - the dread of checking a balance, the embarrassment of asking a friend to split a bill - rather than the rational 'I want lower fees' that customers actually say out loud.
Calm
Meditation and sleep app (USA, founded 2012)Shows the danger of trusting the Says quadrant alone. Users say they want to 'meditate more' and 'be mindful,' but their Does and Feels quadrants reveal the actual job: getting to sleep at 1am while doom-scrolling. Calm leaned into Sleep Stories because the behaviour, not the stated aspiration, pointed there.
Patagonia
Outdoor apparel (USA, founded 1973)An Empathy Map where the Thinks quadrant explains a contradiction the Says quadrant hides. Customers say they buy for quality and durability, but the unspoken thoughts - guilt about consumption, a desire to signal values - are what make 'Don't Buy This Jacket' land instead of backfire.
The 4 Layers, One By One
Each one answers a specific question - here is how to fill it in, and how to tell a sharp answer from a lazy one.
1. Says
What does this person actually say out loud, in their own words, about this situation?
Direct quotes - verbatim if you have them. What the customer voices in interviews, support tickets, reviews, sales calls. The most reliable quadrant because it's observable, but also the most performed: people say the socially acceptable thing.
I just need it to work, I don't have time to read a manual. Your competitor was cheaper but I didn't trust them. Real sentences a person uttered, kept in their voice, not paraphrased into marketing-speak.
The customer wants a seamless, intuitive experience. Nobody says that. That's your slide deck talking. If it doesn't sound like a sentence a human would speak at a kitchen table, it doesn't belong in Says.
2. Thinks
What occupies this person's mind about the situation that they would never say out loud?
The unvoiced thoughts - the doubts, the calculations, the things people are too polite or too self-conscious to admit. This quadrant is mostly inference, so it's the one most in need of evidence. The gap between Says and Thinks is the most valuable real estate on the map.
What if my boss thinks I overpaid? Am I the only one who finds this confusing? Everyone else seems to get it. Private anxieties that drive the decision but never make it into the testimonial.
The customer thinks our product is great. Wishful projection, not inference. If your Thinks quadrant flatters you, you filled it with hope. Real Thinks notes are uncomfortable to read.
3. Does
What does this person physically do - the observable actions and behaviours in this situation?
Actions you can watch, not motivations you imagine. What the person does before, during, and after - the tabs they open, the workarounds they build, the people they ask. Behaviour is honest in a way that stated preference rarely is.
Opens three competitor tabs, reads reviews on their phone in bed, screenshots the pricing to send to a colleague, abandons the cart twice before buying. Each one is a thing you could film.
The customer evaluates options carefully. That's a summary, not a behaviour. Does should read like a stage direction: specific, physical, watchable - not an adjective dressed up as an action.
4. Feels
What is this person feeling, and what specifically triggers each emotion?
Emotional states - but always paired with the trigger, because a feeling without a cause is useless to act on. Frame it as emotion + because. Anxious, relieved, suspicious, proud: name the emotion, then name what set it off.
Anxious because the price felt high with no clear refund policy. Relieved because a real human answered the chat in under a minute. Emotion plus trigger gives the team something to actually fix.
Happy. Frustrated. Satisfied. One-word feelings with no cause attached. They describe a mood, not a moment, and they give the team nothing to design against.
Origin & Lineage
The Empathy Map was created in the mid-to-late 2000s at the design consultancy XPLANE, where founder Dave Gray refined an earlier 'big heads' sketching technique into a structured four-quadrant tool. It was first widely published in the 2010 book Gamestorming: A Playbook for Innovators, Rulebreakers, and Changemakers by Dave Gray, Sunni Brown and James Macanufo, and spread through Stanford's d.school curriculum and design-thinking workshops worldwide. In 2017 Gray released an updated Empathy Map Canvas, built with input from Alex Osterwalder (of the Business Model Canvas), which added goals, numbered the sections to suggest a sequence, and moved Think and Feel to the centre to sharpen the line between observable behaviour and internal mental states. Gray's stated motivation for the update was that the original concept had 'gotten lost in translation' as degraded versions proliferated online.
Critics
The most common criticism is that an Empathy Map filled without real research is just confident guesswork - all assumption, no evidence - that gives a team false certainty about a customer it has never spoken to. Practitioners also point out that the quadrants overlap awkwardly: the same note often fits both Says and Thinks, and Thinks blurs into Feels, so a weakly facilitated map ends up fuzzy and duplicated. A third objection is simple reductiveness - four boxes can't hold context, environment, goals, and social pressure, so the map flattens a human into stickies. The honest defence is the one Dave Gray and NN/g both make: treat empty quadrants as a research backlog, tag every inferred note as an assumption, and never mistake the map for the interviews it's supposed to summarise.
How To Build It
A workshop flow that produces a usable v1 in a day - with the right people in the room, or just you and a Selfstorming strategy session right here.
Decide your starting point
You don't need a wall of sticky notes and a free afternoon to begin. Right here on Selfstorming you can generate a first-draft Empathy Map in minutes from what you already know about your customer. Treat that draft as a hypothesis, not a conclusion - then run it through the steps below to pressure-test every quadrant against real evidence. Workshop-from-scratch and AI-draft-then-verify are both valid; most teams move faster starting from a draft and then hunting for what's wrong with it.
Pick one person and one situation
An Empathy Map of 'our users' covers nobody. Choose a single archetypal person in a single moment - 'a first-time buyer at the checkout,' not 'customers.' The same human Thinks and Feels completely different things at onboarding vs at cancellation.
Gather the raw material before you draw
Pull interview transcripts, support tickets, reviews, sales-call notes, session recordings. The map is a sorting exercise - if you have nothing to sort, you'll fill it with invention and call it insight.
Fill Says and Does first - they're observable
Start with the two quadrants you can source from evidence. Quote people verbatim in Says. Describe watchable behaviour in Does. Anchoring on the observable keeps the inferred quadrants honest.
Infer Thinks and Feels, but tag every guess
These two are mostly inference, so mark each note evidence or assumption. An unlabelled assumption in the Thinks quadrant is how a team convinces itself it understands a customer it has never met.
Hunt the Says-Thinks gap
Put the two quadrants side by side. Where someone says one thing but the evidence suggests they think another, you've found the most useful line on the page - that gap is usually where the real product opportunity lives.
Treat empty quadrants as a research backlog
If a quadrant is thin or blank, don't fill it to look complete. Write the question you can't answer yet and book the interview. The gap is the map earning its keep.
Turn the map into one design decision
A finished Empathy Map that changes nothing was decoration. End every session by naming one thing you'll do differently - a copy change, a feature cut, a support tweak - because of what the map revealed.
How This Framework Compares
| Aspect | When It Works | When It Doesn't |
|---|---|---|
| Best for | Aligning a team on who they're designing for early in a project. Synthesising messy qualitative research into one shared, scannable picture of a customer in a specific situation. | Quantitative segmentation, market sizing, or any decision that needs statistical confidence. The Empathy Map is qualitative scaffolding, not a data model. |
| Output | A one-page, four-quadrant artifact - Says, Thinks, Does, Feels - around a single person in a single moment, with each note traceable to evidence. | A polished persona document with names, photos, and demographics, or a full end-to-end journey. Those are downstream artifacts the map feeds, not the map itself. |
| Time to complete | 20 to 60 minutes per map once the research is gathered. The structure is so simple the time goes into sourcing notes, not drawing boxes. | Multi-week research programmes with recruiting, fieldwork, and analysis. The map is the synthesis step at the end of that, not the whole study. |
| vs Customer Persona | Empathy Map is faster, narrower, and behaviour-first - one person, one situation, four quadrants. Great as the raw input you build a persona from. | A Customer Persona consolidates across many people into a reusable archetype with goals, demographics, and channels. Use the persona when you need a durable reference, not a single-moment snapshot. |
| vs Value Proposition Canvas | Empathy Map captures the whole emotional and behavioural picture of a person without forcing it toward a product fit yet. Better for open exploration. | The Value Proposition Canvas zeroes in on jobs, pains, and gains and maps them directly to your offer. Use it when you're ready to connect customer needs to product features. |
| vs Jobs To Be Done | Empathy Map is a snapshot of a person's inner and outer state in a moment - rich on emotion, light on causality. Good for building team empathy fast. | Jobs To Be Done asks what progress the customer is hiring the product to make, and is stronger on causal 'why.' Use JTBD when you need the underlying motivation, not the surface state. |
Frequently Asked Questions
What is an Empathy Map?
An Empathy Map is a four-quadrant tool for capturing what a specific person Says, Thinks, Does, and Feels in a specific situation. It turns scattered qualitative research - interviews, support tickets, reviews - into one shared, scannable picture of a customer that a whole team can align on. Crucially, it's a synthesis tool, not a research method: it organises insight you already gathered and shows you where your understanding has gaps.
Who created the Empathy Map?
The Empathy Map was created at the design consultancy XPLANE in the mid-to-late 2000s, refined by founder Dave Gray from an earlier sketching technique. It was popularised in the 2010 book Gamestorming by Gray, Sunni Brown and James Macanufo. Gray released an updated Empathy Map Canvas in 2017, developed with input from Alex Osterwalder, that added goals and re-sequenced the sections.
What's the difference between the Says and Thinks quadrants in an Empathy Map?
Says is what the person voices out loud - verbatim quotes from interviews, calls, or reviews. Thinks is what occupies their mind but they won't say, often because of politeness or self-consciousness. The two frequently overlap, and that's fine, but the real value is in the gap: when someone says one thing and the evidence suggests they believe another, you've usually found the product opportunity.
How is an Empathy Map different from a Customer Persona?
An Empathy Map is a snapshot of one person in one moment, organised by Says/Thinks/Does/Feels and built directly from research. A Customer Persona consolidates many people into a durable, reusable archetype with a name, goals, demographics, and channels. The Empathy Map is often the raw input you build a persona from - faster and narrower, where the persona is the polished, long-lived reference.
Does an Empathy Map need real user research?
Yes, and skipping it is the single biggest failure mode. An Empathy Map built from imagination is confident-looking fiction that gives a team false certainty. Source every note from an interview, a ticket, a review, or a recording, and tag inferred notes (mostly in Thinks and Feels) as assumptions. If a quadrant has nothing real to put in it, that empty space is the map telling you which research to do next.
How many Empathy Maps should we make?
One per person per situation that matters. The same customer thinks and feels completely different things at sign-up than at renewal or cancellation, so a single map covering 'our users' across 'their whole experience' ends up describing nobody. Scope each map tightly to one archetype in one moment, then make as many as the journey requires.
Does the Empathy Map work for B2B?
Yes, and the Thinks quadrant does extra work in B2B because buying decisions carry career risk. The unvoiced 'what if this makes me look bad to my boss?' rarely shows up in a sales call but heavily drives the decision. Mapping a specific role in a specific buying moment - 'a mid-level manager evaluating a new tool' - surfaces the political and emotional pressures that a feature comparison sheet completely misses.
Sources & Further Reading
Related Frameworks
Value Proposition Canvas
Two halves, one rule. The customer profile (a circle) maps Customer Jobs (what they're trying to get done), Pains (what gets in the way), an
Jobs To Be Done
Jobs To Be Done says customers hire products to make progress in a situation, not because of who they are demographically. The mechanism is
Bullseye Customer
The Bullseye Customer is three concentric rings of audience. The core is your sharpest, single best-fit early customer - the beachhead you w
Customer Persona
A Customer Persona compresses real research into one archetypal person you can design and write for. Five parts: Profile (who they are in on