Product decisions — Feedback Nexus

This record makes the choices behind the prototype reviewable. It does not report customer evidence or delivery commitments.

Evidence board over an automated roadmap

Alternatives considered: an unstructured inbox, automatic ranking and roadmap generation, or an evidence board with an explicit planning decision. Chosen compromise: preserve individual quotes and account context, expose editable inputs, and let the PM record the final state and rationale. The score is directional support, not a recommendation that applies itself. Cost: more review time and no single objective priority. Reconsider when: sessions show reviewers cannot explain the score, need fields not shown, or prefer a simpler evidence view.

Unique accounts when grouping

Alternatives considered: merge duplicates, count every comment as reach, or retain every signal while showing distinct accounts. Chosen compromise: retain originals and use unique affected accounts to avoid inflation. Individual signals keep their scores; group scores are not added. Cost: reviewers must distinguish related comments from reach. Reconsider when: participants cannot form a decision from grouped evidence or another impact unit proves more meaningful.

Individual score over a group rank

Alternatives considered: calculate a group score from the union, add the scores of linked signals, or keep each source record’s score unchanged. Chosen compromise: the prototype leaves F01 and F02 as individual comparisons and uses the four-account union only to inspect reach. Cost: the reviewer must synthesize related evidence instead of receiving a single group rank. Why: a union says who is represented, not how impact, confidence, or effort should combine. Reconsider when: research shows reviewers can explain a defined group-level unit without mistaking it for a source record or automatic decision.

Simulated tag suggestions with review

Alternatives considered: no suggestion, automatic labels, or a visibly simulated editable suggestion. Chosen compromise: show the suggestion and make acceptance or correction an explicit review action, while allowing a planning state without a prerequisite tag review. Cost: optional review adds triage time and has no model-accuracy claim. Reconsider when: users ignore the tag, mistake it for evaluated AI, or consented evidence justifies separately evaluated assistance.

Local prototype boundary

Alternatives considered: connect live systems, build collaboration, or keep browser-local storage. Chosen compromise: local storage, export, and recovery make the sample inspectable without customer data. Cost: no shared record, permissions, or production workflow. Reconsider when: validated demand supports explicit privacy, access, retention, integration, and operational decisions.

Decision reversal rules