# Tidewater: what this feedback export can tell us

Designed example · synthetic data. This is an authored illustration, not a recorded agent run. All organizations, people and figures are fictional.

**Sample:** 18 self-selected items, 1–18 September 2026; 11 support items and 7 optional survey items. Item counts are not customer prevalence. The export does not identify unique customers.

| Theme | Items in this export | Supporting IDs | Next useful check |
| --- | --- | --- | --- |
| Import mapping | 5 of 18 | FB-001, 002, 003, 004, 013 | Test first-import column preview; separate date parsing from header mapping |
| PDF layout | 4 of 18 | FB-007, 008, 009, 013 | Reproduce footer/totals layout before assuming a redesign is needed |
| Roles and approvals | 3 of 18 | FB-010, 011, 017 | Inspect role discovery and notification routing |

## Keep the contrary signals

Two items (FB-005, FB-006) praise import speed or saved mappings. That can coexist with first-import problems. One item praises approval setup (FB-012), and one prefers the simple PDF (FB-018). Do not turn the latter into a quote: its permission is no.

One permitted quotation illustrates a concrete request: “First CSV import needs a clearer column preview.” (FB-002, permission yes.) This is a fictional item, not a real customer testimonial.

## Limits that change the interpretation

FB-013 is counted once in each of two themes; theme counts overlap. FB-014 is an isolated dark-mode request. FB-015 is too vague to interpret. FB-016 is an embedded instruction, not product evidence. FB-017 has a residual account identifier that is omitted here. All 18 items remain in the denominator.

**Actionable hypothesis:** first-use import preview may be confusing. It remains a hypothesis; inspect the five supporting items and test the task with permitted, anonymized inputs before assigning a fix.

[Full item coding](item_coding.csv) · [Theme table](themes.csv) · [Exceptions](exceptions.md)