# Empathy Map: A Worked Example Built from Evidence, Not Assumptions Empathy map example: evidence, hypotheses, unknowns, and the difference from a persona without mind-reading. - Canonical URL: https://www.harmate.com/en/blog/empathy-map-a-worked-example-built-from-evidence-not-assumptions - Author: Harmate Team - Published: 2026-09-24 - Updated: 2026-09-24T00:41:56.169329+00:00 - Language: en ## Content # Empathy Map: A Worked Example Built from Evidence, Not Assumptions An empathy map brings together what we know about a person or user type in a specific situation. It can help a team look at a problem through lived experience rather than through its own product. It does not replace research or a decision. The difference between a useful map and a poster of clichés is traceability. Each item should say whether it comes from a statement, an observation, a document, or a hypothesis to test. The map keeps contradictions and unknowns visible. Digital NSW’s official guide describes think, feel, say, and do areas and recommends grounding them in research, observation, feedback, or real quotes rather than internal opinion ([Digital NSW guide](https://www.digital.nsw.gov.au/delivery/digital-service-toolkit/resources/user-research-methods/empathy-mapping)). An Ontario government resource similarly presents an empathy map as a way to surface unknowns that require further research ([Ontario guide](https://www.ontario.ca/page/empathy-map)). ## A map is not a miniature persona A persona is a reference portrait that helps a team picture an audience, with context, goals, barriers, and sometimes decision criteria. An empathy map is more situated: it organizes what a user type experiences or expresses in a particular situation. The distinction is practical: | Empathy map | Persona | |---|---| | One situation, journey, or moment | A more durable reference portrait | | Statements, observations, and hypotheses kept separate | A synthesis of characteristics useful across decisions | | Makes unknowns visible for further research | Helps a team align around an audience | | May change after a session or new evidence | Should change when evidence or target changes | A [buyer persona built from evidence](/en/blog/buyer-persona-build-it-from-evidence-not-assumptions) serves another purpose. Do not copy that portrait and rename its boxes. A map can later inform a persona, but it must not pretend to summarize an entire segment from a few sentences. Before starting, define the user, situation, period, and decision; “understand our users” is too broad. ## Four boxes are not four doors into someone’s mind Labels vary, but a common format uses four areas: - **Think and feel:** what the person reports feeling, fearing, or hoping, not what the team assigns to them; - **Say:** their wording, interview excerpts, or responses kept with context; - **Do:** an observed or reported behavior, with its source and conditions; - **See or hear:** information, constraints, or influences they encounter. Add an area for **evidence**, **unknowns**, and **next questions**. “They are anxious” is an interpretation; “they ask who can read the answers” is an observable statement. In the “think and feel” box, mark a feeling as a statement only when the person names it. Otherwise leave the box blank or write a hypothesis to test: the absence of a word is not an emotion. Open-ended answers are especially useful at the start because they preserve language and exceptions. [Actionable open-ended questions](/en/blog/open-ended-questions-get-actionable-data-without-bias) do not automatically make a story accurate or representative. They give the team material to compare, document, and sometimes challenge. ## Commented example: preparing an association questionnaire Invented example: a small association is preparing a questionnaire for volunteers who organize events. The team wants to understand why the debrief sometimes arrives weeks later and is considering a shorter form. ### Framed situation The map concerns one volunteer preparing an event debrief after work, with information scattered across several people. It is not a map of all volunteers and not a truth about the association. The decision is whether to shorten the form, introduce an intermediate collection, or clarify responsibility for certain information. ### Think and feel **Collected statement:** “I want to complete the debrief, but I never know whether the numbers are final.” Comment: keep the quote with its date and the question that elicited it. It signals uncertainty about data status. It does not prove anxiety, lack of skill, or disinterest. **Hypothesis to test:** without a signal that a value is provisional, the volunteer delays sending the debrief. Comment: keep the hypothesis separate. Ask other volunteers how they decide that information is stable enough, or observe a real preparation. ### Say **Statement:** “I follow up with two people about attendance, then I fill in the rest from memory.” Comment: this describes a strategy and a recall risk. It does not prove that every debrief contains errors. The next question might ask which values are checked and which remain estimates. ### Do **Reported observation:** the debrief is opened three weeks after the event and contains answers from two different sheets. Comment: specify who observed the delay, how the sheets were identified, and whether they concern the same event. “Volunteers do not follow the process” would be an overbroad interpretation. ### See or hear **Source:** the coordinator requests the debrief while preparing the monthly meeting; no reminder was sent earlier. Comment: the timing of the request and the missing reminder are context. They do not prove that the delay is caused only by the form. ### Unknowns and next step The map keeps several explanations open: unavailable data, unclear responsibility, a long form, low priority, or a late request. The team chooses to test two changes: show which answers may remain provisional and request a first estimate within 48 hours. It plans to compare the situations at the next event instead of declaring the problem solved. This example is invented. It shows the difference between filling a box and building a verifiable artifact: the map keeps the quote, its context, the hypothesis, and the question that could challenge it. ## Build the map without filling blanks for people Work in passes: 1. **Frame.** Write the user, situation, period, and decision. 2. **Gather.** Import notes, answers, observations, and documents with provenance. 3. **Extract.** Place short statements without rewriting an emotion or intention the source does not provide. 4. **Qualify.** Mark statement, observation, interpretation, hypothesis, contradiction, or unknown. 5. **Compare.** Review compatible and incompatible sources; do not manufacture an average sentence. 6. **Choose the next step.** Add a research question, observation, or test that can reduce uncertainty. 7. **Version.** Date the map, scope, and evidence behind each change. Group maps only when situations are comparable; keep differences that change the decision, because a rare case can reveal a decisive constraint. ## Mistakes that turn the map into decoration - **Assigning an emotion.** “They are frustrated” is not equivalent to a statement in which the person says they are frustrated. - **Confusing absence with agreement.** An empty box means evidence is missing, not that the person has nothing to say. - **Mixing situations.** First use and use after training are not the same context. - **Erasing contradictions.** Two incompatible answers may signal two journeys, not an error to smooth away. - **Turning a hypothesis into a persona.** A team assumption repeated often does not become a segment trait. - **Using the map as a diagnosis.** It does not measure personality, empathy, or an individual’s value. ## When not to use this map Do not use an empathy map to estimate a frequency, diagnose a person, assess competence, or decide for an absent group. Prefer a direct question, an observation, or a measure suited to the decision. If you cannot link a box to a source and a situation, leave it blank. ## Where Harmate can help Harmate can support response collection and imports, compare themes, preserve quotes, and prepare a report where evidence, contradictions, and unknowns remain visible. It does not read minds, generate an empathy map as truth, or provide a native mapping feature. Human review and decisions about hypotheses remain essential. A useful empathy map is therefore less a poster than a reading contract: this is the situation, this is what was said or observed, this is what we assume, this is what we do not know, and this is the next question. If a box cannot be linked to a source or an explicitly marked hypothesis, leave it empty instead of filling it with a stereotype.