Build a customer evidence map for your next product decision

Write the proposed decision in one sentence and identify the customer claim it depends on. “Improve the quoting experience” is an ambition. “Change how exceptions are approved because account managers cannot tell who owns the next action” contains an explanation that can be investigated.

Separate claims that require different evidence. Difficulty finding an owner, disagreement about responsibility and slow approval may occur together, but they are not the same problem. A clearer interface might help one without resolving the others.

Choose the claim that most affects the next commitment. Keep the map focused on that decision.

An evidence map should preserve differences between inputs:

• Observation: Something seen in actual customer work or recorded in product activity, with its context and limitations.
• Reported experience: What someone said happened or mattered to them. Record who the perspective represents without assuming it describes everyone.
• Interpretation: Your team’s explanation of what those observations or reports mean.
• Simulated hypothesis: A possibility explored through an AI-generated perspective that needs an appropriate real-world check.

A product event can show an action occurred without explaining why. An interview can supply context without establishing how widespread a behavior is.

Keep the map small enough to review together:

1. Claim: The customer explanation your proposed decision relies on.
2. Source: Where the supporting material came from, when it was collected and which type of input it is.
3. Context: The role, company situation and task the material concerns.
4. Alternative explanation: Another account that could fit the available information.
5. Decision exposure: What could go wrong if you act and the claim is mistaken.
6. Next check: The smallest useful investigation that could reduce the relevant uncertainty.

Link to the underlying material where your team is authorized to access it. A concise summary should not erase conditions or disagreements that affect its meaning.

Illustrative scenario, not a customer result.

A product team serves mid-sized wholesale distributors in Denmark. It is considering a clearer approval handoff in an existing quoting tool, focusing on account managers and sales-operations leads.

Suppose its research material includes a report of slow approvals and a record showing a quote remained open. The team might interpret both as evidence that ownership is unclear. But the same material could also fit a different explanation: the quote was waiting for external pricing information.

A Virtual Customer conversation can help the team explore questions about these possibilities in the chosen business context. It cannot determine which explanation caused the actual delay.

The map therefore keeps the report, the recorded event and the simulated hypotheses separate. The next check is to reconstruct a recent delayed quote with the people involved, including the information they had and the actions they were waiting for. That investigation can inform whether the proposed handoff change addresses the problem.

Different commitments warrant different levels of support. A reversible prototype used for learning is different from removing a workflow that customers depend on. Record that difference before treating all unresolved questions as equally urgent.

Ask what a finding would change. If the team would make the same decision regardless, either the question is not decision-critical or the proposed investigation is poorly matched to it.

Avoid assigning a universal confidence score to mixed inputs. A large pile of repeated opinions can still leave the core explanation untested. Identify the missing observation or comparison directly.

When the team decides, record the action, its scope and the uncertainty it accepts. Note what would trigger reconsideration, such as a conflicting customer example or an unexpected result after a bounded test.

Preserve disagreement rather than editing the map to make the final choice look inevitable. Date the material so that assumptions tied to an older workflow can be revisited when the product or customer situation changes.

The map makes your reasoning inspectable. It does not transform limited inputs into representative evidence, and simulated hypotheses remain hypotheses until independently investigated. Its practical value is showing what supports the next step and what still needs to be learned.