Research public discussions before choosing a guide topic

Define a discoverable question and a stopping boundary

Write the uncertainty first: “What makes people unsure which reporting period a screenshot describes?” is more useful than “Prove everyone needs my recap.” Define the language, topic, public visibility, dates inspected and time available. A 25-minute discovery session can produce candidate questions; it cannot establish saturation, prevalence or the needs of everyone who uses LinkedIn.

Use an account and interface you are allowed to use. This guide proposes manual reading, not scraping, bulk copying, private-group access or outreach. Search ranking, account context, deleted posts and terminology affect what you can see. Save a small source reference and a relevant paraphrase rather than accumulating profile identities or whole comment histories.

Build and record searches you can actually inspect

LinkedIn documents keyword or conversational searches, a Posts category and category-dependent filters. Its Boolean help documents uppercase AND, OR and NOT, quotes and parentheses. It does not support wildcard asterisks; plus and minus are not officially supported. Record the actual surface and filters you see instead of promising identical behavior in every account or search mode.

Our constructed query plan starts with "reporting period", then tests ("reporting period" OR "date range") AND screenshot, and finally a separately recorded German query such as "Berichtszeitraum". Those are proposed inputs, not searches executed for this guide. A zero-result narrow query calls for a documented revision, not a conclusion that the problem does not exist. Keep each query distinct from the sources it happens to return.

Read context and deduplicate origins

The fictional ledger has six candidates: R1 is an original period question, R2 a repost of R1, R3 an unrelated job advertisement, R4 a second original question about comparison windows, R5 a private discussion that cannot be inspected legitimately here and R6 praise without a task. Retain R1/R4 as two question-bearing origins; mark R2 duplicate, R3 irrelevant, R5 excluded-access and R6 non-question. Six candidates do not become six unmet needs.

For each retained item, read the question, relevant reply and surrounding qualification. Distinguish the author’s own problem from a rhetorical example or a quotation of someone else. Preserve contrary answers and already available solutions. Record the inspected source date and visibility; public availability alone does not license screenshots, names or republishing a whole thread. Use a neutral local reference when private notes suffice.

Resolve three discovery decisions

R-A — Constructed amplification: three queries return R1 and its repost R2. Decision: associate all observations with the same original question rather than count six independent signals. Remaining uncertainty: how many other relevant discussions were invisible. Outcome: one traceable origin, no prevalence estimate.

R-B — Constructed access gap: a search snippet mentions the exact topic but the original requires private membership. Decision: exclude the unread body and preserve the access limitation; do not invent its conclusion from the snippet. Outcome: an unresolved candidate, not supporting evidence.

R-C — Constructed buyer inference: R4 asks how two reporting windows differ. Decision: draft a comparison-window explanation hypothesis. Do not add the author to a prospect list or infer a purchase from the question. Outcome: a possible editorial task that still needs suitable evidence and review.

Turn a pattern into a bounded editorial decision

A useful note says: “Two inspected original questions concern period interpretation; draft a worked comparison and check neighboring guides before commissioning it.” A weak note says: “The market demands a paid dashboard.” Include what the proposed guide would let a reader do, existing answers, contradictory context and what evidence could change your decision.

Stop at the declared boundary and report what was actually inspected. Additional queries can be a new documented session when justified; silently extending until a preferred story appears undermines the record. Combine this discovery with other legitimate evidence rather than treating public search as the entire audience. No automatic persona or model-generated comment supplies another real observation.

Keep a small, useful research record

The packet provides the six-candidate fixture, complete cases and a blank record for question, scope, query/filter state, source/context, inclusion reason, duplicate origin, contrary evidence, proposed task and uncertainty. Its local checker verifies the declared ledger arithmetic only. No real discussion, person, interview, demand, search account or reader study was inspected by the fixture.

Primary help and government user-need guidance were read on 1 November 2026. The latter concerns service design; its evidence-versus-assumption distinction informs this original editorial exercise. Search availability is not a truth, rights or privacy clearance. Keep completed notes private and retain only what the task requires.

1. Record the task, scope and actual authority.
2. Inspect inputs and evidence state at the actual source.
3. Work through the full proposal with dependencies and unknowns.
4. Record the decision, owner and next permitted step.
5. Verify actual results separately from plans and constructed examples.