Context
A LinkedIn content team currently needs to decide which publishing patterns to repeat. It needs an account-level view of recent public posts so recommendations are based on comparable evidence rather than one standout or weak post.
The user's question： “How should this LinkedIn account post for higher engagement, and can high readership be copied directly?”
Analyze 95 LinkedIn posts for a content-strategy shift and replicable engagement patterns
Compare content types across the returnable post history instead of copying one viral post, and keep missing impressions from being treated as reads.
Answer： Across 95 returnable posts, App rankings averaged 15.3 likes and hashtag-only posts averaged 2.1. Readership cannot be verified because impressions were null on all 50 recent posts.
- Case source
- Authorized user case
- Research category
- Social research
- Executed
- Last updated
- Platforms
- Tools used
$research-social-signals
What is public
The user authorized publication of redacted engagement metrics, content types and the research conclusion. Account name, profile and company details remain redacted.
The actual data
Bars start at zero. This is an account-internal comparison, not a LinkedIn benchmark and not readership.
Subject
95 returnable LinkedIn posts
- Returnable sample
- 95
- Ranking avg likes
- 15.3
- Hashtag-only avg likes
- 2.1
- Impressions
- null
Pagination chain ended
7 posts
14 posts
Reads unverifiable
Average likes by content type in the recent 50 posts
Labeled by publish date · zero baseline
Growth cases6.6
DTC / brand marketing4
Hashtag-only posts2.1
App rankings15.3
How the data changed the judgment
Initial concern
Copy the highest-readership post directly.
What the data showed
Impressions were missing. The gap is between image-led data rankings and hashtag-only posts.
Revised judgment
Retest rankings and single-app cases; stop hashtag-only posts. This is an account ranking, not a platform algorithm rule.
How to reuse this method
- 01Confirm available metrics and treat missing fields as unknown.
- 02Follow pagination instead of treating page one as the full history.
- 03Compare by content type, not only the top post.
- 04Adjust new posts by engagement per day.
- 05Keep temporal alignment separate from causation.
Review the execution details
Method and scope
- Pull the latest posts and pin down metric boundaries：Fetched the first 50 posts via get_linkedin_user_posts. Impressions were null on all 50, so likes, comments, and reposts were the available interaction metrics; readership was not treated as a verifiable result.
- Follow pagination tokens to complete the returnable history：Used the page-one pagination_token to retrieve 45 page-two posts, then used its token for page three. Page three returned zero posts, ending the available chain. The final sample is 95 posts from 2024-05-17 through 2026-08-07.
- Profile the account and the engagement overview：Confirmed the account profile (a market researcher with several hundred followers and 500+ connections; name and company redacted) and first computed the recent-50 totals: 295 likes (median 4, max 46), 22 comments on 7 posts, 15 reposts on 7 posts; 34/50 posts with images and 34/50 long-form posts. The 95-post sample was then used to check strategy evolution.
- Rank top posts and break down by content type：Ranked the 50 posts by likes+comments+reposts and grouped them into four content types: App rankings (7 posts, avg 15.3 likes, 36% of likes), growth case studies (10, avg 6.6), DTC/brand marketing (13, avg 4.0), and hashtag-only posts (14, avg 2.1).
- Analyze the time trend and content shape：Found a visible turning point: June 2026 had 98 total likes and 12.2 average likes, up from 17 and 4.2 in May; July reached 13.7 average likes. Rankings began in June as content moved from hashtag-only/generic marketing to data-driven rankings and growth cases. These variables moved together and establish temporal association only.
- Adjust engagement for post age：Calculated engagement velocity as likes, comments, and reposts divided by days since publication, preventing recent posts from being penalized for their shorter accumulation window. A single-App case post led the sample at 1.33 interactions per day; the measure remains an account-internal comparison.
- Extract post text and deliver the complete Markdown report：Extracted the full redacted text of key high-engagement posts and checked their data-led openings, tension, structured points, takeaway, and question CTA. The delivered Markdown report includes a topic matrix, replication formula, three drafts, a 95-post appendix, early-versus-recent comparison, monthly trend table, and Bottom 15 counterexamples.
Research conversation
User
@skill:Social Signal Retrieval Analyze the posting activity of this account: https://www.linkedin.com/in/xx. The goal is to find a replicable method for posts with high readership.
SignalDig research agent
Understood. I’ll first run a deep comparison: high- versus low-engagement post characteristics, opening patterns, and publishing cadence. I’ll then deliver a complete quantitative Markdown report.
SignalDig research agent
The data is clear. I’ll add one essential control: engagement velocity (engagement per day), so a newly published post is not misclassified as weak simply because it has had less time to accumulate interaction.
SignalDig research agent
The data is sufficient. Next I’ll extract the full text of key posts for a post-by-post breakdown and adaptable examples; account identity, links, and company details will remain redacted.
SignalDig research agent
The analysis is complete. I’m upgrading the earlier findings to a quantitative report based on all 50 posts; before delivery, I’ll read the existing report and replace it with the updated Markdown version.
SignalDig research agent
The full quantitative Markdown report was delivered, based on 50 posts from 2025-11-07 through 2026-08-07. It includes two validated higher-engagement content lines, an engagement-velocity adjustment, topic priorities, publishing observations, three adaptable drafts, and a post-level appendix.
User
Add more page-level data.
SignalDig research agent
I’ll use the pagination_token returned on page one to fetch pages two and three, then compare older material with recent posts instead of treating the latest 50 as the account’s entire history.
SignalDig research agent
Page two returned 45 posts and a further token. Page three returned zero posts, so this returnable pagination chain is exhausted. The combined sample contains 95 posts from 2024-05-17 through 2026-08-07, spanning 812 days.
SignalDig research agent
I’ll update the Markdown report and monthly trend chart with the complete sample, adding an early-versus-recent comparison, a monthly trend table, and a Bottom 15 counterexample section. Content shape and engagement both change around June 2026; that is temporal alignment, not proof that a single series caused the growth.
Execution evidence
01
The execution established scope, metric definitions, and missing fields
Page one returned 50 posts from 2025-11-07 through 2026-08-07. Each was checked for likes, comments, reposts, publish time, text, and image fields; impressions were null on all 50. Following the page-one token returned 45 posts from May 2024 to October 2025; page three returned zero. The report therefore uses likes+comments+reposts and engagement per day, while explicitly treating readership as unverified.
02
Full pagination revealed a measurable shift in content shape
Across the 45 early posts, average likes were 0.20, 88.9% had zero likes, 95.6% were hashtag-only, mean text length was 139 characters, and image rate was 6.7%. Across the 50 recent posts, the corresponding figures were 5.9, 8.0%, 30%, 1,368 characters, and 68%. The comparison shows engagement and content format changing together; it does not isolate any one field as the cause.
03
Fifty posts exposed a clear content-type gap
App-ranking posts averaged 15.3 likes across seven posts and carried 36% of all likes, while fourteen hashtag-only posts averaged 2.1. The comparison identifies a relative high-engagement line inside this account, not a platform benchmark.
04
The time series corrected the earlier single-post conclusion
June 2026 reached 98 total likes and 12.2 average likes; July average likes reached 13.7. May had 17 total likes and 4.2 average likes. The “Global Mobile App Revenue Top 30” ranking series began in June, and its first post earned 46 likes. With missing impressions and uncontrolled variables, this supports retesting the content line—not a readership, reach, or causal claim.
05
Grouping and age adjustment produced an actionable, bounded replication target
All eight high-engagement posts included images and had a 2,025-character median length; the 24 low-engagement posts had a 1,034-character median, and 14 hashtag-only posts generated just 30 interactions combined. After age adjustment, a single-App case post ranked first at 1.33 interactions per day. The report prioritizes App rankings and single-App cases and recommends stopping hashtag-only posts; this is account-level evidence, not a LinkedIn algorithm rule.
06
The deliverable is traceable to post-level data rather than a generic recommendation
The updated Markdown report includes a 95-post appendix ranked by total engagement and engagement per day, content-type summaries, early-versus-recent comparison, a monthly trend table, Bottom 15 counterexamples, key-post text breakdowns, a data anchor → tension → structured points → takeaway → CTA formula, a topic matrix, and three drafts requiring human rewriting and fact checks. It records missing impressions, pagination scope, and the non-causal nature of the comparison.
Limits and what this does not prove
- Impressions (the "readership" the user wanted to replicate) were null on the initial 50 posts, so engagement metrics remain a proxy, not reach; extra pages do not turn that proxy into readership.
- The 95 posts are complete only for this retrieval’s returnable pagination chain, not necessarily the account’s permanent full LinkedIn history; deletion, visibility, and endpoint scope can change what is returned.
- 4 posts had missing like_count and were treated as missing, not zero.
- This is a single-account snapshot; patterns are not a platform-wide benchmark.
- Engagement velocity depends on capture date and publish time. Without impressions, traffic-source, or audience-composition data, it cannot establish true reach efficiency.
- Weekday and time-of-day observations come from a non-random account sample; topic, image choice, text length, and account growth were not controlled. The early-versus-recent comparison also cannot turn correlation into causation.
Continue with a related research task
Compare the closest verified executions before generalizing the conclusion.
Tools used
$research-social-signals