Empathy Maps · UI/UX Atlas

Key takeaways

An empathy map structures research output into four evidence-typed quadrants — Says and Does are observational; Thinks and Feels are inferences — keeping them distinct prevents the say/do gap from contaminating synthesis.
Build collaboratively from raw primary research, not from team assumptions; a map sourced from assumption is a liability that mimics the look of rigor without any of the value.
The artifact's job is to feed downstream synthesis — personas, journey maps, opportunity statements — not to stand alone; without a clear next step, it collects dust.
Triangulate qualitative map findings against behavioral and quantitative data wherever possible; convergent evidence dramatically raises confidence in a pain point or need.
Version and date every map tied to a specific research round; a stale empathy map can be more misleading than having none at all.

The full lesson

Research data is only as useful as the shared understanding it creates. Teams often collect solid interview recordings, diary entries, and observation notes — then fail to act on them because the insights live in one researcher’s head or a spreadsheet nobody reads.

An empathy map fixes that. It is a fast, visual synthesis artifact that externalizes what you learned about a user or user segment. Every stakeholder can see it, challenge it, and build from it together.

This lesson covers the anatomy of a modern empathy map, how to build one rigorously from real research, where it fits in the broader synthesis workflow, and the common habits that turn a powerful tool into decorative wallpaper.

What an Empathy Map Actually Is

An empathy map is a four-quadrant synthesis artifact. Each quadrant captures a different dimension of the user’s experience:

Says — direct quotes or paraphrases from research sessions; verbatim language matters
Does — observable behaviors, actions, sequences, and workarounds you witnessed or were told about
Thinks — inferred beliefs, assumptions, questions, and mental models; things the user may not say aloud
Feels — emotional states, anxieties, motivations, and moments of frustration or delight

Many formats also include two derived zones — Pains and Gains — placed below the quadrants. These summarize core tensions and desired outcomes. They are not a fifth input. You draw them as conclusions after filling the four quadrants.

The Say/Do Gap and Why It Matters Here

Modern research treats behavioral data as more reliable than self-report. When a participant says “I always read the documentation first” but your session recording shows them skipping straight to trial and error, the Does quadrant wins.

An empathy map that conflates these two layers produces false confidence. The resulting artifact looks rich but is contaminated by aspiration, social desirability bias, and post-hoc rationalization.

The fix is disciplined sourcing. Every sticky note or entry in the map should cite its evidence type: observation, direct quote, or researcher inference from tone or body language. This practice makes the map auditable and keeps synthesis conversations honest.

When to Build an Empathy Map

Empathy maps are generative synthesis tools. They belong in the sense-making phase after qualitative data collection, not at the start of a project. They are not a substitute for research — they are a way to process and share what research produced.

Primary use cases:

After user interviews or contextual inquiry sessions, before moving to persona or journey map creation
When onboarding new team members to an existing research base — the map is faster to absorb than raw transcripts
At the start of a design sprint to align a cross-functional team on a shared user understanding before ideation
When a research repository exists but the team is still arguing about “what users want”

When an empathy map is the wrong tool:

As a substitute for actual research (building one from team assumptions is a proto-persona exercise, not synthesis)
As the terminal synthesis artifact — it should feed into personas, journey maps, or opportunity statements, not sit alone
For quantitative findings — frequency counts, task-completion rates, and statistical patterns belong in other artifacts

Building an Empathy Map from Real Research

Preparation: What You Need Before You Start

Gather your raw data: transcripts, annotated session recordings, observation notes, diary study entries. You need primary sources — not a previous researcher’s summary, which already filters out nuance you may need.

Recruit the right builders. The most valuable empathy maps are built collaboratively with 3–6 people who were present in the research or have deep domain context. A lone researcher building a map and presenting it as a finished deliverable misses the synthesis value. The act of debating where a note goes — or whether something belongs in “Thinks” versus “Feels” — is where understanding sharpens.

Define scope before you start. Are you building one map per participant, one per behavioral segment, or one aggregate map? For 5–8 qualitative sessions, start with per-participant maps, then merge them into a composite that represents a behavioral archetype. For larger research programs with 20 or more sessions, affinity diagramming across the full dataset is a more appropriate first step before collapsing into empathy map clusters.

The Build Process

Step 1 — Data download (15–20 minutes)

Each team member writes one observation per sticky note, drawing directly from the raw data. Enforce a constraint: use the participant’s words for Says, and observable actions for Does. No interpretation yet.

Step 2 — Sort and cluster (20–30 minutes)

Place notes on the four-quadrant board. Physical boards, Figma, FigJam, Miro, and Mural all work well. Resist the urge to force balance. If you have twelve Does observations and two Says, that uneven distribution is itself a finding. Cluster duplicate or closely related observations.

Step 3 — Inference layer (15 minutes)

Now fill Thinks and Feels from the clustered evidence. These cells require interpretation. Explicitly label them as researcher inferences and note which Says and Does observations they derive from.

Step 4 — Pains and Gains synthesis (10–15 minutes)

Pains are unmet needs, frustrations, obstacles, and fears. Gains are desired outcomes, aspirations, and measures of success. Both should be tension-filled and specific. “Wants things to be faster” is too vague to act on. “Needs to complete the approval workflow during a between-meetings window of under 4 minutes” is actionable.

Step 5 — Challenge and validate (15 minutes)

Read the map back against your raw transcripts. Remove anything unsupported. Flag any inference where the evidence is thin. Mark items that appeared in only one session versus those that recurred across multiple participants.

Digital vs. Physical Facilitation

Remote teams commonly build empathy maps in FigJam or Miro with color-coded sticky notes per participant. This works well but introduces a specific risk. Virtual boards lower the friction for adding notes, which means low-quality or unsourced entries accumulate faster. Counter this with explicit labeling conventions — a color or tag that marks the note type (verbatim quote, observation, inference) — and a dedicated pruning pass before sharing the map more widely.

Physical maps built on a whiteboard or large paper have the advantage of creating shared spatial memory. Team members remember “that anxiety cluster in the bottom left.” The disadvantage is they do not persist well. Photograph and digitize immediately, and tag every cluster with a summary label so the photograph is scannable.

The Anatomy of a High-Quality Map

The difference between a map that drives decisions and one that collects dust usually comes down to a few attributes:

Attribute	High-quality map	Low-quality map
Evidence sourcing	Every entry linked to a session, observation, or quote	Unsourced; built from team assumptions
Says/Does distinction	Observational and reportable data kept separate from inferences	All quadrants treated as equal-confidence input
Specificity	Concrete, behavioral, contextual	Abstract, generic, aspirational
Pains/Gains	Derived from the quadrants, tension-filled, actionable	Copied from a brief or strategy doc
Scope	Single behavioral archetype or defined user segment	”All users” or undefined audience
Updatability	Dated, versioned, tied to specific research rounds	No provenance; unclear if still current

Source every note directly from raw research. Label the evidence type — verbatim quote, observation, or researcher inference. Use specific, contextual language (names of tasks, tools, environments) rather than abstract generalizations. Build collaboratively so the map encodes distributed team knowledge. Date and version the artifact so its freshness is always apparent.

Don't

Fill an empathy map from team assumptions and present it as if it were grounded in research. Mix verbatim participant quotes with researcher paraphrases without distinguishing them. Build a single generic map for “all users” — segment first, then map. Treat the map as a finished product; it should feed downstream artifacts, not stand alone. Copy Pains and Gains from a product brief rather than deriving them from the quadrant evidence.

Empathy Maps and the Broader Synthesis Stack

An empathy map does not stand alone. It sits in a synthesis pipeline between raw data and higher-order artifacts:

Raw research (transcripts, recordings, notes) → Empathy maps (per-participant or per-segment; what they say, do, think, feel) → Personas (named behavioral archetypes with goals, contexts, mental models) → Journey maps (how the persona experiences a process end-to-end, with emotional arc) → Opportunity statements or How Might We prompts (design challenges extracted from pain points)

The empathy map is the first filter. It takes voluminous, unstructured research data and reduces it to a structured set of human signals. Personas consume that output and add narrative and decision-making utility. Journey maps then put the persona in motion across a temporal experience.

Teams that skip the empathy map phase and try to build a persona directly from raw transcripts typically produce underspecified archetypes. Those archetypes collapse under the first design review challenge: “how do we know this user actually feels that way?” The empathy map is the evidence trail that answers that question.

Common Failure Modes

The Assumption Map

The most common failure: a team with no time for research fills an empathy map in a workshop using their own intuitions. The artifact looks identical to a research-backed map but is worthless — or worse, actively harmful. It gives assumption-based decisions a veneer of user-centeredness.

The fix is not to skip the map. Label it explicitly as a proto-artifact with named assumptions that must be validated. Schedule a revisit date. Treat every quadrant entry as a hypothesis.

The Static Artifact

Empathy maps built once and never updated become a liability as your user population, product, or context shifts. A 2023 map describing how users navigate a desktop-only workflow is a poor guide to a 2026 mobile-first redesign. Attach empathy maps to the research rounds that produced them. When a new research round contradicts a prior map, update it visibly rather than letting the stale version persist in shared drives.

The Four-Equal-Boxes Trap

Many teams treat all four quadrants as equally important inputs with equal evidence requirements. In practice, Says and Does should always be richer than Thinks and Feels. The latter require inference that is only as reliable as its observational basis. A map with abundant Thinks and Feels but sparse Says and Does is a warning sign — it means the team is projecting more than observing.

Over-Granularity vs. Vagueness

Sticky notes that are too granular (“clicked the blue button”) add noise without synthesis. Notes that are too abstract (“wants a good experience”) add nothing actionable. Target the middle level: a specific observation that implies a pattern (“re-reads confirmation email three times to verify the order was placed — doesn’t trust the on-screen confirmation”).

Relationship to Mixed-Method Research

Modern synthesis practice does not treat qualitative and quantitative research as competing paradigms — it triangulates across them. An empathy map is primarily a qualitative synthesis tool, but its findings should be checked against behavioral data when available.

If an empathy map shows that users feel anxious about data loss, validate that inference. Do analytics show high rates of repeated save actions, or elevated drop-off after errors? If a pain in the map (“hard to find previous orders”) aligns with a high support ticket volume on that topic, confidence in that finding increases substantially. This cross-validation step is what separates research-backed synthesis from storytelling.

From Map to Action

A well-built empathy map should be directly usable in the next phase of work. Test yours against these criteria before calling it done:

Could a new team member read this map and make a design decision they could not have made before?
Are the Pains specific enough to turn into “How Might We” prompts without additional clarification?
Could you defend every cell entry with a specific research citation if challenged?
Is the scope — user segment, context, and research round — stated explicitly?

If the answer to any of these is no, the map needs another synthesis pass — not more content, but tighter, better-sourced content.