Experiment framework Test seven organic LinkedIn text posts over four weeks, with no more than two posts weekly. Compare each post’s organic impressions at seven days. Audience: US freelance graphic designers managing three to six client projects and browsing to understand the offer. Facts: downloadable pack of six editable project checklists; six editable documents plus a setup guide; USD 24 one-time. English text, maximum 900 characters, no hashtags, owner review required. Resources: one owner, three hours weekly, existing editor. Photos and demonstration are optional and will remain unused to keep format constant. Estimates: production and administration budgets below. Hypotheses: angle, text presentation, and reply timing may affect impressions. Primary metric: organic impressions per post at seven days. Diagnostic metrics: organic impressions at 24 hours; reader comments at seven days, excluding owner comments; substantive reader questions at seven days. Data quality: reporting access is a fictional scenario assumption. Audience qualification and non-follower status are unavailable. No publication is authorized. Exclude paid boosts, artificial engagement, spam, invented testimonials, unsupported results, and third-party media. Baseline and comparability Existing value Period Denominator Source No baseline Preceding 30 days No organic posts Supplied brief Organic impressions unavailable before testing Seven days per planned post One post Fictional reporting setup; verify before execution Post 1 establishes an initial measurement basis. Posts 2, 4, and 6 provide local references for their paired treatments. Impressions count exposures, not unique people. Do not substitute reach, views, engagement rates, profile actions, or follower changes. Use the same elapsed observation windows for every post. Paid distribution is excluded; follower groups cannot be separated with the supplied data. Hypothesis portfolio Hypothesis Proposed mechanism Observable signal Counter-observation Data risk Angle: project context versus pack contents Mentioning concurrent projects may make the offer more immediately relevant Post 3 exceeds post 2 in seven-day impressions Equal or lower impressions; questions suggest confusion Different publication dates; relevance cannot be verified from impressions Presentation: short lines versus paragraph Short lines may make supplied facts easier to scan Post 5 exceeds post 4 in seven-day impressions Equal or lower impressions; no useful qualitative difference Date effects; one treatment cannot confirm readability caused the change Interaction: earlier reply opportunity Earlier substantive replies may sustain an existing conversation Post 7 exceeds post 6, with eligible early comments and replies recorded No improvement despite an actual early reply opportunity No early comments means the treatment was not exercised; comment quality varies These mechanisms are hypotheses, not claims about LinkedIn’s algorithm. Seven-post matrix All posts use the same verified offer facts, price, neutral interaction question, and text-only format. References repeat one approved master post exactly. No full post copy is included. No. / week Hypothesis Topic angle Format Entry idea Core benefit Interaction impulse Creation* Controlled variable 1 / W1 Initial reference Pack contents One paragraph Introduce the downloadable pack Six editable documents and setup guide Ask what needs clarification about the contents 35 min Establish master 2 / W1 Angle reference Pack contents One paragraph Same approved entry Same Same 20 min Exact master 3 / W2 Angle treatment Managing concurrent projects One paragraph Address designers handling three to six projects Same; no efficiency promise Same 35 min Opening angle only 4 / W2 Presentation reference Pack contents One paragraph Master entry Same Same 20 min Exact master 5 / W3 Presentation treatment Pack contents Short lines Identical master wording Same Same 35 min Line breaks only 6 / W3 Interaction reference Pack contents One paragraph Master entry Same Same 20 min Standard reply checks 7 / W4 Interaction treatment Pack contents One paragraph Master entry Same Same 35 min Additional early reply check *Creation excludes review and administration, budgeted in section 9. Design: one initial reference plus three reference/treatment pairs: 2–3, 4–5, and 6–7. Each treatment changes one essential variable. Publication date, audience overlap, and repeated wording are confounders. Keep topic constant within the presentation and interaction pairs. Later replication is required before declaring any winner. Platform-appropriate implementation Use only LinkedIn text posts: Maximum 900 characters, including spaces and line breaks; no hashtags. References and angle treatment use one paragraph. Presentation treatment changes only line breaks; retain identical words and punctuation. No audio, images, video, subtitles, or alternative text are required because no media is attached. Use plain English, readable sentences, and no decorative Unicode formatting. Owner checks character count, factual accuracy, price, exclusions, and the single intended variable. The public sample has not been inspected. Do not describe its contents, reproduce it, or claim evidence from it. Distribution protocol Proposed schedule assumption: Tuesday and Thursday at 12:00 America/New_York in weeks 1–3; Tuesday at that time in week 4. These are administrative windows, not verified audience activity times. Time and date differences remain confounders. For every post: Publish only after owner approval and verification of actual reporting access. Check comments at approximately 24 and 72 hours. Answer genuine questions using supplied facts; do not invent missing answers. No unsolicited promotional comments, tagging campaigns, crossposting, reposting, or requests for coordinated engagement. Do not add standalone owner comments. For post 7 only, add a comment check 15 minutes after publication and reply to genuine comments already present. Keep the later checks unchanged. Record whether an early reply actually occurred. Cap interaction at 20 minutes per post. Log unanswered questions if capacity is exhausted. No real publication occurs under this brief. Measurement plan Event or metric Denominator Measurement time Observation window Source Organic impressions, diagnostic One post Publication +24 hours First 24 hours Scenario reporting setup Organic impressions, primary One post Publication +168 hours First seven days Scenario reporting setup Reader comments One post +168 hours First seven days Manual count; exclude owner comments Substantive reader questions One post +168 hours First seven days Manual classification Early interaction exposure Eligible comments present at 15 minutes Post 7 +15 minutes First 15 minutes Owner log Record publication timestamp, collection timestamp, text version, approval, and owner reply times. For each pair, report reference and treatment impressions, absolute difference, and percentage difference relative to the reference. If reference impressions are zero, the percentage difference is undefined. Report comments and questions as counts and, where impressions exceed zero, per 1,000 impressions. These ratios are diagnostic, not unique-person engagement rates. Log outages, edits, unusual mentions, missed checks, crossposting, or accidental paid boosts. Mark affected comparisons as compromised rather than silently pooling them. Decision rules Apply these rules after all seven-day windows close: Continue testing: a treatment has higher seven-day impressions than its reference, the manipulation occurred, and no material confounder explains the result. Replicate before confirmation. Maintain format: presentation treatment shows no clear advantage, or the simpler reference is easier to produce within capacity. This is an operational choice, not proof of equal performance. Discard topic angle: do not carry the tested opening into the next round if it has lower impressions and reader questions indicate misunderstanding. Discard provisionally; one comparison cannot establish a general failure. Unclear: results conflict across observation windows, reporting is missing, a comparison is compromised, or post 7 receives no eligible early comments. Review qualitative questions for offer understanding. Few or no comments provide little qualitative evidence. No statistical certainty or minimum sample size is claimed. Do not select a winner from the highest-impression post alone; assess its paired reference, treatment exposure, and confounders. Production handover Weekly capacity budget, in minutes: Week Creation Review/accessibility Publishing Interaction Measurement/decision Total 1: posts 1–2 55 20 10 40 25 150 2: posts 3–4 55 20 10 40 25 150 3: posts 5–6 55 20 10 40 25 150 4: post 7 and earlier results 35 10 5 20 40 110 5: post 7 measurement tail 0 0 0 0 45 45 Each week stays within 180 minutes. Week 5 includes the final seven-day reading and decision; publishing ends in week 4. Before publishing — owner role: Verify reporting access, approve facts and wording, check accessibility and length, confirm only the intended variable changed, and record the planned window. Immediately after — publisher role: Record actual timestamp and approved version; confirm organic publication; log deviations and follow the interaction protocol. During evaluation — analyst role: Collect comparable windows, exclude owner comments, calculate paired differences, check confounders, and apply the predefined rules. One owner fills all roles. Next-test conclusion: select exactly one hypothesis for replicated follow-up. Choose the angle if it shows an interpretable paired improvement; otherwise choose presentation if it does; otherwise choose early interaction if it was exercised and improved. If none qualifies, replicate the angle hypothesis to resolve uncertainty. This priority rule selects one next test without declaring a winner.