Sample walkthrough — Feedback Nexus

This is a public walkthrough of the fictional Northstar sample. It uses only browser-local data; no feedback is sent anywhere. Start from the original fixtures: choose Reset sample data, read the confirmation, and reset it. Reset replaces local browser edits.

1. Open Feedback inbox and search for row. You should see F01 — Show exactly where an import failed and F02 — Retry failed rows without starting over. F01 starts with Alder Labs, Finch Studio, and Cedar Systems; F02 starts with Pineworks and Alder Labs.
2. Select F01 and F02, then choose Group selected (2). The records stay visible as separate evidence: neither quote nor account list is deleted. Open F01’s detail panel. Its affected-account line reads the unique four-account union: Alder Labs, Finch Studio, Cedar Systems, Pineworks. Do not add the two displayed counts; Alder Labs appears in both source records. If a saved record contains Alder Labs, alder labs, or another case/whitespace variant, it still contributes one account. The first source spelling is shown so the evidence remains recognizable, while the directional score and export use the normalized unique set.
3. Clear the search, then compare that import-recovery evidence with F04 — Save the weekly operations report. F04 remains separate and starts as Inbox. This is a deliberate PM tradeoff: reach is a directional input, while the original quote, segment, confidence, and effort remain available to challenge the comparison.
4. Return to F01 by selecting its record, then inspect the suggested Onboarding tag. You may accept it, edit it, or leave it unreviewed. A nonblank edit saves when you leave the tag editor; a blank draft shows guidance and preserves the last valid saved tag. Tag review is optional: it supplies context and does not block a planning decision.
5. Change F01 confidence to 40%. Its individual score becomes 1.6: three original accounts × impact 4 × 0.4 ÷ effort 3. The four-account group union is evidence context; it does not alter F01’s individual score. The score is not a delivery commitment or an automatic recommendation.
6. Enter a decision rationale and move F01 to Planned. Open Product shortlist to see the individual Planned record, then use Export shortlist. The interface reports 1 planned feedback record exported. The CSV includes Planned individual records only, orders multiple records by their directional score, and labels account reach as Original record accounts so a linked group is not mistaken for a scored aggregate.
7. Select Undo to restore the last saved change. Reset sample data opens a confirmation dialog before replacing browser changes; that replacement can also be undone. If the browser reports incompatible saved data, edits remain session-only and the stored payload is preserved until you explicitly confirm Reset sample data. The fallback fixtures are isolated from the reference sample, and padded or unresolved saved group labels are rejected instead of being guessed.

The walkthrough leaves one question deliberately open: whether preserved evidence and a transparent directional score help real PMs more than their current workflow. The proposed study tests that before adding automation, collaboration, or integrations.