Why Journey Maps and Empathy Maps Fail to Drive Product Decisions

Something changed in how good research teams work this year, and it has less to do with new tools than with where the fights actually happen. The methods got sharper. So did the politics around them. A journey map or an empathy map does not just describe users anymore. It moves power around the room, and people feel that. Let me catch you up.

The deep cut

• Research artifacts move power, not just understanding. A retail bank's journey map turned into a turf war the moment it exposed duplicate work.
• A sparse map is a research gap, not a draft. Nielsen Norman warns an empty empathy quadrant means you need more data, not better guesses.
• Reframe resistance as business risk. "How much to build this three times?" opened doors at the bank that experience talk never did.

The map that earns its keep

An empathy map is a simple grid: Says, Thinks, Does, Feels, with the user in the middle. The Nielsen Norman Group frames it as a way to pool what your team knows and spot what it does not. The honesty rule is the whole thing: every sticky note traces back to a transcript line or a session note. If you cannot point to the source, label it an assumption.

The test is brutal and useful. Can't fill a quadrant? That is a signal to go get more research before you design further. The map that gets presented once and filed away is worthless. The value shows up later, when a design decision is contested and you ask what the map actually supports.

When "Anxious" is useless

Specifics are the job. The group warns that a quote like "I want something reliable" could come from any user, so it gives you nothing. "I screenshot the confirmation page every time, because I don't trust that the email will actually arrive" tells you exactly where trust broke.

Same rule in the Feels quadrant. "Anxious" on its own is dead weight. "Anxious: can't tell whether the payment went through" points straight at a fix. Watch the contradictions too. A positive action next to a negative feeling usually means your user did not say something out loud. Your job is to chase that down, not smooth it over.

The map that redistributes power

Here is where the methods meet the politics. A large retail bank had three product teams, each with its own acquisition journey. Someone from Design mapped them together, and the map showed customers hitting wildly inconsistent experiences while the same capability got built three times. The reaction was immediate. Every team felt its roadmap was under threat, and the talk flipped from "how do we improve this" to "who gets control."

The team that got traction stopped selling experience quality and reframed it as cost: how much are we spending to build the same thing three times? Finance became an ally. The lesson repeats down the org chart. A service designer found that technical stakeholders resisted a journey pilot because the content was theirs and their standing came from owning it. What resolved it was private trust, not a better slide.

Tension you work through instead of around

Some conflicts do not have a clean answer. In a child support research project, Julia Beard and her coauthors heard one parent say payments pushed them into debt and the next parent say the same structure left them short on food and rent. Both true. Picking a side would hurt someone real, so they built a Range of Tolerance: not the solution everyone loves, but the one every voice can live with without harm.

That discipline connects to a plainer warning from Emily Cohen: the real risk is not making assumptions, it is forgetting they are assumptions and treating them as fact. Her fix is cheap. Ask more questions. The cost of a few extra questions is far lower than deciding on what you think you know.

Three questions for your team

• Pick one empathy map from a recent project. Can every note on it trace back to a real research artifact, or are we designing off memory?
• When our last journey or audit met resistance, was the objection really about method, or about who stood to lose ownership? What would the cost framing have been?
• What is one assumption about our users we have been treating as fact, and what would it cost us to actually test it this quarter?