What goes in each quadrant
Says is their actual words. Quote them where you can — a real sentence carries nuance that a paraphrase sands off, and it keeps the map honest about what you heard versus what you concluded. Thinks is what occupies them that they did not say out loud, whether because it was too obvious to mention, too awkward, or not yet conscious. That gap between says and thinks is often the most useful thing on the map.
Does is observable behaviour: the steps they take, the tools they open, the workarounds they have built. Watch for the workarounds especially — an elaborate one is a user telling you precisely where your product fails, in a form they will never write in a survey. Feels is the emotional state, named plainly: anxious about being blamed, embarrassed to ask again, relieved when the day ends.
Pains are the frustrations, risks and obstacles in their way. Gains are what a good outcome actually looks like from where they sit — which is frequently not the outcome your roadmap is optimizing for.
One person, not an average
The most common way an empathy map goes wrong is by describing a segment rather than a person. 'Small business owners feel time-poor' is a demographic statement; nothing on the map contradicts it, so nothing on the map can teach you anything. Name a specific person, or a specific composite drawn from a handful of interviews you actually did, and write about them.
The second failure mode is filling it from imagination. An empathy map built by four people in a room who have never spoken to a user is a record of the team's assumptions — which is a legitimate thing to make, as long as you label it that way and treat every entry as a question to test rather than a finding. Marking your assumptions distinctly costs nothing and stops the map from hardening into false confidence.
Using the map after you make it
Read across the quadrants for contradictions. Someone who says the process is fine, does an elaborate manual workaround, and feels anxious about it every Friday is a much richer brief than any of the three observations alone. Contradictions are where the real design problems hide.
Then convert what you found. A pain that appears for several people becomes a problem statement; a gain nobody on the team had considered becomes a design goal; an empty quadrant becomes an interview question. If a map does not change what you build or what you ask next, it was a workshop activity rather than research.
Empathy map, persona, or journey map?
A persona is a durable profile of a user type. An empathy map is a snapshot of one person's inner and outer experience, usually built as an input to a persona or straight after a round of interviews. A journey map is the same person over time, stage by stage, and answers where in the process the trouble happens.
In practice: empathy map to understand the person, journey map to find the moment, then a prioritization tool to decide what to do about it. All of them save to the same canvas here.