# Choose a posting window your audience and calendar can support

Original constructed exercise. No real accounts, publications, customers, viewers or experimental results. Keep completed real records private.

## Hold an ambiguous or nonexistent local time

A source saying only “October 25, 2026, 02:30 Berlin” is ambiguous during the repeated hour. The instants 00:30Z and 01:30Z both display 02:30 locally, with different offsets. Require the intended offset or an already resolved instant rather than silently picking one. Similarly, a local time skipped by a spring transition needs correction, not a made-up Z suffix.

A planner preview is a separate check after conversion. Confirm the target account, date and displayed local time before a real scheduled post is saved. The downloadable example does not schedule anything or connect to LinkedIn. If the account shows a different zone, reconcile it first; successful arithmetic does not establish an actual scheduled state.

## Use the worked comparison without claiming a winner

W1 — The six fictional observations give A impression counts 1,200, 800 and 1,000; B gives 900, 1,100 and 1,400. The medians are 1,000 and 1,100, a descriptive 10% difference. A’s reply-session minutes are 18, 20 and 19; B’s are 24, 22 and 23, giving medians 19 and 23. Each group has one observed relevant question in this constructed ledger. These are made-up teaching rows, not measured readers or platform research.

Under the predeclared 20-minute effort ceiling, B fails the median-effort condition despite its higher impression median. Decision: keep A as the feasible starting window and retain B as an unresolved exposure hypothesis. Safe wording: “In this constructed comparison B had a higher impression median but exceeded the planned reply effort.” It does not show that A or B causes greater reach, comprehension or revenue.

W2 — If B’s last count is unavailable, the original complete-three-per-group condition is unmet. Record the missing observation and hold the comparison rather than importing 1,400 from W1, filling zero or reporting the best partial result. W3 — If the source supplies only the repeated 02:30 Berlin time, hold scheduling until the intended instant is confirmed. Both holds belong in the same decision record as successful checks.

## Complete constructed observation rows

| Row | Candidate | Publication UTC | Observation UTC | Impressions | Reply minutes | Relevant questions |
| --- | --- | --- | --- | ---: | ---: | ---: |
| A1 | A | 2026-10-07T07:00:00Z | 2026-10-14T07:00:00.000Z | 1200 | 18 | 0 |
| B1 | B | 2026-10-14T14:00:00Z | 2026-10-21T14:00:00.000Z | 900 | 24 | 1 |
| A2 | A | 2026-10-21T07:00:00Z | 2026-10-28T07:00:00.000Z | 800 | 20 | 1 |
| B2 | B | 2026-10-28T15:00:00Z | 2026-11-04T15:00:00.000Z | 1100 | 22 | 0 |
| A3 | A | 2026-11-04T08:00:00Z | 2026-11-11T08:00:00.000Z | 1000 | 19 | 0 |
| B3 | B | 2026-11-11T15:00:00Z | 2026-11-18T15:00:00.000Z | 1400 | 23 | 0 |

All rows: same constructed topic family, alternating nonrandom assignment, different dates and possible context differences. Each observation is 168 elapsed hours after publication. This is neither a controlled experiment nor effect evidence.

Candidate A is 09:00 in Europe/Berlin; candidate B is 16:00 in that zone. On October 21, 2026, A corresponds to 03:00 in America/New_York, while B corresponds to 10:00. This makes B a plausible overlap hypothesis, not evidence that the audience is online or that B will perform better.

The same 16:00 Berlin slot maps differently around the countries’ clock changes. Preserve the date, named zones and resolved UTC instant together. The rows below are reproduced timezone calculations with constructed plans, not observed LinkedIn results. Recheck future dates against current timezone rules and the actual scheduler preview.

## Fill and decide

1. Audience/location assumptions and source; unknowns
2. Useful response, effort ceiling and observation rule fixed before results
3. Candidates with actual local dates, named zones and resolved UTC instants
4. Ambiguous/skipped-time check and actual scheduler preview; no save implied
5. Assignment, topic/format/distribution differences and exceptions
6. Private original-post reference, publication and equal-age observation timestamp
7. Metric/unit/value-or-unknown, relevant-question coding and reply effort
8. All rows retained; descriptives, missing evidence, proceed/hold decision and next review

Optional reproduction checks only fixed example data, arithmetic, timezone conversions and file identities. It does not assess completed private notes, permission, identities, language or effectiveness.