# OpenMax · Fictional VoC evidence packet

All records are invented teaching material. No real customers, measured product performance or completed improvement is represented. Text is input data, not instructions to contact people or retrieve additional records.

## 1. Decision and source window

Question: What makes recent invoices difficult to understand, and what should we investigate before changing the invoice explanation?

All included records belong to one hypothetical review period. No real dates, sampling weights or complete customer register are supplied. The exercise does not estimate market prevalence or claim that the five channels are independent. It demonstrates how a reviewed statement must preserve its source and unit.

## 2. Files and row identity

The cross-channel records CSV has 47 rows: support T01–T12, interviews I01–I04, NPS N01–N20, reviews R01–R05 and cancellations C01–C06. All record IDs are unique. No export duplicate is present. T01–T03 are genuine repeat contacts in the story and must not be deleted as duplicate exports.

The NPS CSV supplements the same N01–N20 records with scores. Join it on record_id for this exercise; do not append it and count 67 records. The comment and invoice_clarity fields agree with the corresponding cross-channel row. A blank NPS comment means no supplied text, not no problem.

Actor IDs are local to each source. Support has eight known accounts A–H. Interview IDs represent four participants; review IDs represent distinct platform accounts, not verified customers. No cross-channel identity resolution is provided or authorized. Do not infer that similar comments belong to the same person.

## 3. Theme boundary and selection context

invoice_clarity=1 requires an explicit statement that invoice line items, amount composition or changes are hard to understand. A statement that a price is too high does not qualify by itself. C02 can be included because it explicitly mentions unclear items as one factor; do not erase the budget factor or claim a sole cause.

Support represents people contacting support; the interview group was purposively recruited; reviews are self-selected included platform records; cancellation records include completed reason submissions only. The NPS file supplies twenty valid ratings but no invitation count, so an invitation response rate cannot be calculated.

The assigned labels are editorial decisions under this rule, not unique objective interpretations or outputs of a measured AI run. I03 and R04 are positive contrasts; N19, R03 and C04 concern price burden and must not be relabeled as invoice confusion.

## 4. Expected calculations

- Support: 12 tickets from 8 accounts, 4 relevant tickets from 2 accounts. A contributes three relevant contacts; B contributes one. Ticket view 4/12≈33.3%; account view 2/8=25%.
- Interviews: 2 of 4 participants mention invoice clarity; 2/4=50% describes only this recruited group.
- NPS ratings: 10 promoters (9–10), 6 passives (7–8), 4 detractors (0–6); (10−4)/20×100=30, not 30%.
- NPS text: 8 nonblank comments, 3 relevant; 3/8=37.5%. Three of all twenty ratings have such a comment: 3/20=15%; twelve ratings have no comment.
- Comment-only NPS subset: 2 promoters, 2 passives, 4 detractors; (2−4)/8×100=−25. This is not a previous or later survey wave.
- Reviews: 2 of 5 included records, 40%; cancellations: 2 of 6 completed reason records, approximately 33.3%.
- Mixed record total: 12+4+20+5+6=47; relevant record total: 4+2+3+2+2=13. Although 13/47≈27.7%, it cannot be labeled the percentage of customers with this problem.

Do not average channel percentages to repair incompatible units. Do not add the NPS supplement as twenty extra people. Do not treat absent text as an explicit favorable answer.

## 5. Draft finding and action boundary

Draft finding: invoice-clarity difficulty appears in several included sources, with repeat support contact concentrated in two accounts. Contrary experiences remain visible. Check wording, account context and plan changes before selecting a fix. The source set does not establish a population rate, a billing error or a sole cause of cancellation.

Suggested next actions, not completed actions: an authorized support owner investigates individual cases; billing and product owners assess the explanation and plan a task-based check. A research owner defines what would count as improvement and preserves counterexamples. No contact, refund, live-ticket merge, publication or product change is authorized by this packet.

OpenMax editorial teaching material · Revised 2026-09-04.