# Original review worksheets and fictional worked examples

Every account, count, source and result in these worked examples is constructed. These files are not a LinkedIn export, incident dataset, real search result, reader study or customer outcome. The worksheets are blank and perform no account action, upload, save, send or publication.

## Reproduce the arithmetic

Extract the ZIP into an empty working folder. With Node.js 24.21.0 (the version actually checked), run:

```sh
node reproduce-examples.mjs
```

No packages, account, credentials or network access are needed. The command checks the complete constructed inputs and the SHA-256/byte identities of the named files in `file-identities.json`. It does not check an actual platform account or interpret a filled-in private worksheet. Other Node versions were not exercised for this standalone packet.

Expected results:

- `comment-events.json`: 7 displays, 7 fictional accounts, 5 reactions, 20 comments, 25 response events. Repeated comments explain why event count and account count are different. This does not explain every real reaction-only discrepancy or confirm/refute a platform incident.
- `paired-observations.json`: matched reaction-event ratios 12/300 = 4% and 30/900 ≈ 3.33%. Mixing the later numerator with the earlier denominator gives 10%, neither matched observation. These are a separate constructed post B, not a recovered missing historical count.
- `network-shares.json`: outside shares 40% and 80%; the combined weighted share is 480/1,100 ≈ 43.64%, while the unweighted share mean is 60%. These exact inputs are fictional impression events, not new viewers/followers.
- `reach-overlap.json`: three per-post counts sum to 9; the fully specified example union has 5 fictional account IDs. Two other set assignments give unions of 3 and 9 with the same per-post totals. Aggregate counts alone cannot identify which union is real. Estimated production metrics add uncertainty.
- `retrieval-exercises.json`: four fictional decisions distinguish a checked candidate, a liked-but-not-saved item, an archive reference with unavailable content and an unsent draft. No retrieval was executed in a private LinkedIn account.

## Use a blank worksheet privately

- `metric-evidence-en.md` / `metric-evidence-de.md`: preserve metric labels, scope, clocks, observed states and one controlled repeat. A record is not a submitted support ticket.
- `observation-log-template.csv`: a header-only starting point. Preserve blank/loading/error/unavailable separately from an observed zero. No software here parses or validates a completed CSV.
- `retrieval-log-en.md` / `retrieval-log-de.md`: record the actual object, query/filters, source check, result and next step. A missing result does not prove nonexistence; a reference does not prove accessible original text.
- `draft-check-en.md` / `draft-check-de.md`: verify the actual saved/restored state, file version and audience. Draft, scheduled and publicly published are different states.
- `audience-check-en.md` / `audience-check-de.md`: review the native audience and actual media. A setting does not establish permission, confidentiality or complete recall.

Keep private file paths, names, account details, draft text and original screenshots in appropriate private storage. Share only the minimal necessary redacted record. Do not enter real data into these public example files or commit private worksheets to a public repository. Read the corresponding guide's current official sources before acting; controls and supported environments can change. The tools in this packet send nothing automatically.

This packet improves a workflow. Its exercises, file downloads and successful checks count as no real readers, independent reviewers, customers, platform operations or additional original research study.