What Works on LinkedIn.
Every public post from 346 prominent LinkedIn writers, 74,119 in all, collected on 10 September 2026 and measured by length, line rhythm, first line, format, image shape, links, hashtags, timing, cadence, subject, profile and banner. Each band is indexed against the median post and checked per follower, on comments and within the author, so the page says which habits lift a post and which only mark a big account.
The study is liveStudy 002 · Published 9 October 2026
Seventeen sections follow, each with its tables and a short reading. The study names no author and quotes no post: it reports the shapes of posts, never the people who wrote them. Every index is correlational across accounts of very different sizes, which is why each headline figure is shown beside the lens that strips account size out.
Methodology: the population, the measures, the lenses and the limitsMedia enquiries
- Study
- 002
- Profiles
- 346
- Posts
- 74,119
- Learning set
- 39,858 posts, 187 authors
- Collected
- 10 September 2026
- Published
- 9 October 2026
The headline findings
-
Hashtags cost engagement
Posts with no hashtags index 1.05; one to three index 0.82 and more than three 0.84. Comments fall further than reactions: a hashtagged post earns 0.71 of the comment share. The strongest authors hashtag 6% of their posts against 10% for the typical author.
-
The link card is what hurts, not the link
A link card (the article preview) indexes 0.51 and, within the same author, earns 0.52 of their image posts. A URL typed into the text of an image post costs far less (0.88 within the author) and a URL in the last fifth of the text indexes 1.24. "Link in the comments" changes nothing: 0.97.
-
Text-only posts do worst of all
Within the same author, a text-only post earns 0.39 of their image posts, and the text-only band indexes 0.71 across the set. 36,046 of the 39,858 posts in the learning set carry media.
-
Portrait beats square beats landscape
A portrait image indexes 1.13, square 0.93, landscape 0.81 and no image 0.87. Portrait also leads per follower (1.22) and on comments (1.19).
-
The first line is a lever, and it is short
A first line under 40 characters indexes 1.13; over 140 characters, 0.76. Four to seven words is the strongest band (1.11, comments 1.23); over twenty words, 0.82 with 0.49 of the comment share. Two sentences on the first line is the weakest shape on every lens.
-
Long posts are a big-account habit
Over 2,000 characters indexes 1.29 raw, the highest length band, but only 0.88 per follower: it is what large accounts write, not what lifts a post. 700 to 2,000 characters leads per follower at 1.12. Under 300 characters indexes 0.59 on any lens.
-
White space and short lines, every time
Over eight blank lines indexes 1.11; one to three, 0.66. Short lines (under 60 characters) index 1.08 against 0.85 for paragraphs. Under 60 characters before the first break indexes 1.12.
-
The next day beats the same day
A post 20 to 30 hours after the author's last indexes 1.13; under six hours, 0.80; more than eight days later, 0.63. Within the author, a post under twelve hours after the previous one earns 0.82 of their posts twenty or more hours apart. The strongest cadence band is 6 to 7 a week.
-
Afternoons lead, but timing is a weak lever
12 to 17 UTC indexes 1.16 and 14:00 UTC is the peak hour at 1.20; the same author's afternoon posts earn 1.07 of their others. Per follower the afternoon reads 0.93, so the slot is also where the biggest accounts post. Sunday 12 to 17 UTC is the strongest cell (1.34); Thursday is the weakest day (0.93).
-
Most engagement advice is measuring account size
A P.S. indexes 1.19 across authors and 1.70 on comments. Within the author, each post against their own median, the whole close dimension spans 0.99 to 1.04, and a question mark spans 0.99 to 1.01. The habit belongs to big accounts; it does not lift a post.
-
Emoji are not a lever
Three or more emoji index 1.10 and none 0.89, but the strongest authors use them in 66% of posts and the typical author in 62%: the same rate. In the first line an emoji indexes 0.92 and takes 0.82 of the comment share.
-
Reposts earn a fifth; quote posts a twentieth
An original post indexes 1.06, a repost 0.83 and a quote post 0.20.
How to read the numbers
The index is the band's median engagement against the learning set's median. Engagement is reactions plus comments on a post. The learning set is the 39,858 posts from the 187 authors whose own median is 100 or more, and its median post earns 320. A band at 1.13 has a median 13% above that; a band at 0.82 sits 18% below it. A band needs twenty posts to be reported.
Four lenses, because account size is the confound. The raw index counts audience as much as craft: a habit of the largest accounts reads well whatever it does. So each key band is also read per 1,000 followers (median engagement per thousand followers of the author, indexed against 7.2 for the set), on the comments share (comments as a share of all engagement, indexed against 18.8%), and within the author: each post divided by its own author's median, so an index of 1.07 means the band's posts run 7% above what their authors usually get. Where the raw lens and the within-author lens disagree, the within-author reading is the one to trust, and section 5 shows why.
What is not here. No author, handle, post or first line. No reach or impressions, which LinkedIn does not publish. No claim that any band causes anything: a post that breaks every rule on this page can still be the best post of the week if it has a point.
In the tables. Each index cell carries a short bar against the set's median (the tick is 1.00): gold above it, grey below. The strongest band in a table is marked.
Length
Length is where the lenses disagree most, and that disagreement is the finding. Raw, the longest band wins: over 2,000 characters indexes 1.29, the highest reading of any length band, and over 300 words 1.27. Per follower, the same band reads 0.88: the authors who write past 2,000 characters are the accounts with the most followers, and the raw index is counting their audience. The band that leads per follower is 700 to 2,000 characters at 1.12, with 1.13 of the comment share.
The floor is firm on every lens. Under 300 characters indexes 0.59; under 700, 0.75 per follower and 0.67 on comments. The strongest forty authors run a median of 1,162 characters a post against 1,090 for the typical author.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| over 2000 | 1.29 | 5,209 | 410 |
| 1200 to 2000 | 1.07 | 13,269 | 341 |
| 700 to 1200 | 0.98 | 11,917 | 312 |
| 300 to 700 | 0.87 | 6,023 | 279 |
| under 300 chars | 0.59 | 3,440 | 188 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| over 2000 | 1.28 | 6.4 | 0.88 | 19.2% | 1.02 | 5,209 |
| 700 to 2000 | 1.03 | 8.1 | 1.12 | 21.2% | 1.13 | 25,184 |
| under 700 chars | 0.78 | 5.4 | 0.75 | 12.5% | 0.67 | 9,436 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| over 300 words | 1.27 | 7,146 | 405 |
| 200 to 300 words | 1.05 | 11,508 | 336 |
| 100 to 200 words | 0.97 | 13,701 | 310 |
| under 100 words | 0.72 | 7,503 | 231 |
Read the two length tables together and the shape is clear: write enough to make a point (700 characters and up), and do not mistake the long-post index for a rule. A 2,500-character post from a 10,000-follower account is not carrying the audience that makes the raw band look good.
Line rhythm and white space
How a post is laid out moves the needle more than how long it is. Short lines, under 60 characters each, index 1.08; medium lines 0.93; paragraphs over 110 characters a line, 0.85. Both cohorts agree (craft 0.88, domain 0.84 for paragraphs).
Blank lines tell the same story from the other side: over eight indexes 1.11, four to eight 0.90, and one to three or none 0.66. The strongest authors run a median of 12 blank lines a post. What sits above the first break matters too: under 60 characters before the first break indexes 1.12, against 0.94 for 60 to 140. Sentences under eight words index 1.09.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| short lines (under 60 chars each) | 1.08 | 21,743 | 345 |
| medium lines (60 to 110) | 0.93 | 11,908 | 298 |
| paragraphs (over 110 chars a line) | 0.85 | 6,207 | 272 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| over 8 blank lines | 1.11 | 25,210 | 355 |
| 4 to 8 blank lines | 0.90 | 8,823 | 288 |
| 1 to 3 blank lines | 0.66 | 3,676 | 212 |
| no blank lines | 0.66 | 2,149 | 210 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| under 60 chars before the first break | 1.12 | 17,819 | 356 |
| 60 to 140 before the first break | 0.94 | 16,849 | 299 |
| over 210 before the first break (past the fold) | 0.93 | 2,520 | 296 |
| 140 to 210 before the first break | 0.83 | 2,670 | 266 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| under 8 words a sentence | 1.09 | 6,932 | 347 |
| 8 to 12 words a sentence | 1.07 | 14,045 | 341 |
| 12 to 18 words a sentence | 0.97 | 10,815 | 310 |
| over 18 words a sentence | 0.87 | 7,231 | 278 |
| no sentence punctuation | 0.75 | 835 | 238 |
One sentence per line, a blank line between, and a first line that ends before the reader has to scroll. The paragraph post is the weakest rhythm band in both cohorts, which is a measured fact about the feed rather than a taste.
The first line
The first line is the only part of a post everyone sees, so nine things about it were measured, each on four lenses. Length first: under 40 characters indexes 1.13, 40 to 80 1.01, 80 to 140 0.90, over 140 0.76. In words, four to seven is the strongest band on the raw index (1.11) and on comments (1.23); over twenty words indexes 0.82 raw, 0.54 per follower and 0.49 on comments, the single weakest reading in the hook tables.
Shape: one full sentence indexes 1.04; a line that ends open on a colon or an ellipsis 1.06; a fragment 0.99 raw but 1.22 per follower; a question 0.88; and two or more sentences on the first line 0.88 raw with 0.62 of the comment share, the weakest opener shape on every lens. An all-caps word in the first line indexes 0.84 (comments 0.70); an emoji 0.92; a number is not a lever (0.99). A first line that names no subject indexes 1.06 against 0.95 for one that does, on the raw lens; within the author both read 1.00.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| first line under 40 chars | 1.13 | 11,275 | 359 |
| first line 40 to 80 | 1.01 | 18,922 | 321 |
| first line 80 to 140 | 0.90 | 6,292 | 286 |
| first line over 140 | 0.76 | 3,369 | 243 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| 4 to 7 words | 1.11 | 1.15 | 1.02 | 1.23 | 11,345 |
| 1 to 3 words | 1.02 | 1.41 | 1.01 | 1.14 | 1,805 |
| 8 to 12 words | 1.01 | 1.06 | 1.00 | 1.13 | 14,128 |
| 13 to 20 words | 0.93 | 0.90 | 0.99 | 0.90 | 7,710 |
| over 20 words | 0.82 | 0.54 | 0.96 | 0.49 | 4,841 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| ends open (colon or ellipsis) | 1.06 | 0.88 | 1.03 | 1.06 | 3,110 |
| one full sentence | 1.04 | 1.01 | 1.00 | 1.08 | 20,593 |
| a fragment | 0.99 | 1.22 | 1.03 | 1.05 | 8,880 |
| a question | 0.88 | 0.94 | 0.94 | 1.06 | 2,017 |
| two or more sentences on the line | 0.88 | 0.75 | 0.98 | 0.62 | 5,229 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| no all-caps word | 1.01 | 1.01 | 1.00 | 1.03 | 36,341 |
| carries an all-caps word | 0.84 | 0.91 | 0.97 | 0.70 | 3,488 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| no emoji | 1.02 | 1.00 | 1.00 | 1.04 | 34,379 |
| carries an emoji | 0.92 | 1.05 | 1.02 | 0.82 | 5,450 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| no number | 1.00 | 1.02 | 0.99 | 0.97 | 30,247 |
| carries a number | 0.99 | 0.94 | 1.03 | 1.10 | 9,582 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| says you | 1.03 | 0.98 | 0.98 | 1.08 | 9,514 |
| says I or we | 1.00 | 1.08 | 1.04 | 1.07 | 12,781 |
| neither you nor I | 0.99 | 0.98 | 0.99 | 0.92 | 17,534 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| names no subject | 1.06 | 1.11 | 1.00 | 1.08 | 18,830 |
| names a subject in the line | 0.95 | 0.91 | 1.00 | 0.93 | 20,999 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| a negation or contrast | 1.04 | 0.98 | 1.00 | 0.98 | 7,793 |
| no contrast | 0.99 | 1.01 | 1.00 | 1.01 | 32,036 |
| Band | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|
| stands on its own | 1.00 | 1.00 | 1.00 | 1.00 | 37,903 |
| opens with a dangling reference | 0.99 | 1.12 | 1.04 | 1.03 | 1,926 |
Four to seven words, one sentence, no capitals and no emoji. The one surprise is the subject line: across authors a first line that withholds its subject reads higher, which is the curiosity hook at work, and within the author the two shapes are level. The within-author column is the one to trust where they disagree.
Openers and archetypes
Beyond the first line, what kind of post is it? Among openers, a story anchored in time (a day, a year, an age) indexes 1.17; a number 1.12; first person 1.04; a question 0.88.
Classified by archetype, the feed is mostly observation: 55% of posts are an observation or a take, and they index 0.97. A how-to or numbered post (9.1% of posts) indexes 1.13 with 1.33 of the comment share; a contrarian claim 1.09; a personal story (15%) 1.07 raw and 1.07 within the author, the strongest within-author reading of any archetype after the announcement (1.13, on 2.7% of posts). A data drop indexes 0.92 and a question-led post 0.87.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| a story, anchored in time | 1.17 | 1,107 | 372 |
| a number | 1.12 | 1,522 | 356 |
| you | 1.09 | 2,301 | 348 |
| contrarian claim | 1.09 | 2,100 | 348 |
| I or we | 1.03 | 5,461 | 329 |
| another statement | 0.99 | 23,018 | 315 |
| a fragment of three words or fewer | 0.95 | 1,395 | 302 |
| a question | 0.88 | 1,965 | 280 |
| a quotation | 0.80 | 989 | 255 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| opens with a number | 1.12 | 1,628 | 356 |
| opens in first person | 1.04 | 5,249 | 333 |
| opens another way | 1.00 | 29,793 | 320 |
| opens with a question | 0.88 | 3,188 | 280 |
| Band | Share of posts | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|---|
| how-to or numbered | 9.1% | 1.13 | 0.78 | 1.03 | 1.33 | 2,731 |
| contrarian claim | 4.9% | 1.09 | 1.09 | 1.00 | 1.10 | 1,464 |
| personal story | 15% | 1.07 | 1.16 | 1.07 | 1.16 | 4,464 |
| announcement | 2.7% | 1.03 | 1.07 | 1.13 | 0.76 | 809 |
| observation or take | 55% | 0.97 | 1.01 | 1.01 | 0.90 | 16,339 |
| data drop | 9.0% | 0.92 | 0.93 | 1.00 | 1.10 | 2,674 |
| question-led | 4.6% | 0.87 | 0.96 | 0.95 | 1.02 | 1,379 |
Archetypes are assigned by rule from the text, one per post; the rules are in the methodology.
The personal story is the archetype that lifts an author above their own median most reliably, and the question-led post is the one that lands below it on every lens. The how-to earns comments; the announcement earns reactions from the author’s own audience and few comments.
The close
The ending is where this study changed its own mind. Across authors, a post that ends on an ask indexes 1.17 and a P.S. 1.10; on the lenses a P.S. reads 1.19 raw, 1.35 per follower and 1.70 on comments, and ending on a question 1.55 on comments. Those are big numbers and they are the ones most LinkedIn advice is built on.
Within the author, each post against that author's own median, the whole close dimension spans 0.99 to 1.04: a P.S. ending on a statement 1.04, a plain question close 1.03, a P.S. ending on a question 1.02, a plain statement 0.99. Question marks anywhere in the post run from 0.99 to 1.01. The cross-author readings were measuring the size of the accounts that use the habit, not the habit.
| Band | Share of posts | Index (raw) | Index per follower | Within the author | Comments index | Posts |
|---|---|---|---|---|---|---|
| an ask | 11% | 1.17 | 0.83 | 1.01 | 1.04 | 3,404 |
| a P.S. | 11% | 1.10 | 1.16 | 1.02 | 1.63 | 3,145 |
| a statement | 72% | 0.96 | 0.97 | 1.01 | 0.89 | 21,413 |
| a question | 5.8% | 0.90 | 1.41 | 1.03 | 1.34 | 1,735 |
| ends on a list | 0.5% | 0.73 | 1.03 | 1.05 | 0.83 | 163 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| has a P.S. | 1.19 | 9.8 | 1.35 | 31.8% | 1.70 | 4,372 |
| no P.S. | 0.97 | 7.0 | 0.97 | 17.2% | 0.92 | 35,457 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| ends on a question | 1.06 | 10.0 | 1.38 | 29.1% | 1.55 | 3,099 |
| ends on a statement | 0.99 | 7.0 | 0.97 | 17.9% | 0.95 | 36,730 |
| Band | Index within the author | Posts |
|---|---|---|
| a P.S. ending on a statement | 1.039 | 3,553 |
| no P.S., ends on a question | 1.033 | 2,283 |
| a P.S. ending on a question | 1.022 | 819 |
| no P.S., ends on a statement | 0.994 | 33,203 |
| Band | Index within the author | Posts |
|---|---|---|
| one question mark | 1.008 | 10,184 |
| 2 or 3 question marks | 1.001 | 7,882 |
| 4 or more question marks | 1.000 | 4,034 |
| no question marks | 0.994 | 17,758 |
End on sense. A P.S. is worth about a point either way once the author is held constant, and a question mark is not a lever at all. The other fourteen dimensions were not re-read on this question in the same depth; where a cross-author index looks too good, assume account size until the within-author column says otherwise.
Format, images and links
By format, video indexes 1.23 raw but 0.84 per follower and 0.67 on comments; image 1.07 raw and 1.17 per follower; a document (carousel) 0.93 raw with the highest comment share of any format at 1.47; text only 0.58; a link card 0.51, the weakest format on every lens.
Within the author the gaps are wider than the raw bands show: a text-only post earns 0.39 of the same author's image posts (14 wins in 108 authors), a link card 0.52 (7 in 41), and an image post with a URL in its text 0.88 (45 in 119). A URL in the last fifth of the text indexes 1.24, earlier in the text 1.03. Saying the link is in the comments indexes 0.97: the folklore is not supported. A post carrying any link reads 1.18 raw and 0.55 per follower, another big-account habit.
Image shape: portrait 1.13, square 0.93, landscape 0.81, no image 0.87, and portrait leads per follower and on comments too. Four or more images index 1.15 against 1.05 for one.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| video | 1.23 | 5,188 | 391 |
| image | 1.07 | 27,201 | 340 |
| document (carousel) | 0.93 | 2,215 | 297 |
| text only | 0.58 | 3,769 | 184 |
| link card (article preview) | 0.51 | 1,485 | 164 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| video | 1.22 | 6.1 | 0.84 | 12.6% | 0.67 | 5,187 |
| image | 1.06 | 8.5 | 1.17 | 20.9% | 1.11 | 27,187 |
| document | 0.93 | 5.3 | 0.73 | 27.5% | 1.47 | 2,215 |
| text only | 0.58 | 4.1 | 0.57 | 15.9% | 0.85 | 3,766 |
| link card | 0.52 | 2.3 | 0.32 | 10.7% | 0.57 | 1,474 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| video / url in text | 1.52 | 1,151 | 486 |
| image / url in text | 1.30 | 5,475 | 415 |
| document (carousel) / url in text | 1.23 | 620 | 393 |
| video / no url | 1.15 | 4,037 | 366 |
| image / no url | 1.02 | 21,726 | 325 |
| document (carousel) / no url | 0.90 | 1,595 | 287 |
| link card (article preview) / no url | 0.65 | 860 | 206 |
| text only / url in text | 0.58 | 430 | 186 |
| text only / no url | 0.58 | 3,339 | 184 |
| link card (article preview) / url in text | 0.37 | 625 | 118 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| url in the last fifth | 1.24 | 6,634 | 394 |
| url earlier in the text | 1.03 | 1,667 | 330 |
| no url in text | 0.97 | 31,557 | 308 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| does not | 1.00 | 39,387 | 320 |
| says the link is in the comments | 0.97 | 471 | 309 |
| Comparison | Median ratio | Authors where the first wins | Authors measured |
|---|---|---|---|
| the same author's image posts with a URL in the text, against their image posts without one | 0.88 | 45 | 119 |
| the same author's link-card posts against their image posts | 0.52 | 7 | 41 |
| the same author's text-only posts against their image posts | 0.39 | 14 | 108 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| portrait image | 1.13 | 8.8 | 1.22 | 22.3% | 1.19 | 19,681 |
| square image | 0.93 | 9.8 | 1.35 | 18.8% | 1.00 | 4,262 |
| no image | 0.87 | 5.0 | 0.69 | 14.7% | 0.78 | 12,642 |
| landscape image | 0.81 | 5.9 | 0.82 | 14.4% | 0.77 | 3,244 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| 4 or more images | 1.15 | 959 | 367 |
| one image | 1.05 | 23,984 | 335 |
| 2 or 3 images | 1.00 | 2,909 | 319 |
| no image | 0.90 | 12,006 | 286 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| carries a link | 1.18 | 4.0 | 0.55 | 16.4% | 0.87 | 8,294 |
| no link | 0.97 | 8.3 | 1.15 | 19.5% | 1.04 | 31,535 |
Post an image, in portrait. Never a link card. If a link must go in the post, put it at the end of the text; "link in the comments" is a ritual that changes nothing measurable. Video reads well raw and badly per follower, which says the biggest accounts post video rather than that video lifts a post.
What the strongest image posts show
The corpus measured whether a post had an image and what shape it was, not what was in it. So the 94 highest-lifting image posts (one per author, each post against the author's own median) were read by eye and classified. 50% are a photograph of the author; 17% a meme; 11% a text card; 7.4% a screenshot; 6.4% a chart. 61% carry words on the image.
Within that set, the meme lifts most (1.44 against the set's median lift of 7.57 times the author's median) and the photograph of the author least (0.86), though the photograph draws the most comments (1.27) after the screenshot (2.24). Text on the image reads 1.11; an image with no people in it 1.11.
| Kind | Share of the set | Lift over the author’s median | Lift index | Comments index | Posts |
|---|---|---|---|---|---|
| has text: text on the image | 61% | 8.43x | 1.11 | 0.90 | 57 |
| has text: no text on the image | 39% | 5.76x | 0.76 | 1.23 | 37 |
| kind: meme | 17% | 10.92x | 1.44 | 0.52 | 16 |
| kind: text-card | 11% | 7.70x | 1.02 | 0.68 | 10 |
| kind: chart | 6.4% | 5.72x | 0.76 | 1.33 | 6 |
| kind: photo-of-author | 50% | 6.54x | 0.86 | 1.27 | 47 |
| kind: screenshot | 7.4% | 7.13x | 0.94 | 2.24 | 7 |
| people: no people in the image | 31% | 8.43x | 1.11 | 0.90 | 29 |
| people: people in the image | 69% | 7.25x | 0.96 | 1.03 | 65 |
Half of the biggest image wins are the author’s own face, and the face is the kind that lifts least and converses most. The meme and the text card carry the words; the photograph carries the person. A read of 94 posts is a small set and is reported as one.
Numbers, lists, quotes and voice
Posts with six or more numerals index 1.12, three to five 1.04; a percentage or a money figure 1.05. Four or more list lines index 1.10; a quoted passage 1.06.
Voice: "you" four or more times per hundred words indexes 1.08, and "I" under twice per hundred words 1.10 against 1.04 for four or more. A first line that says "you" reads 1.08 on comments; one that says "I" or "we" 1.08 per follower.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| 6 or more numbers | 1.12 | 10,903 | 356 |
| 3 to 5 numbers | 1.04 | 7,456 | 332 |
| 1 or 2 numbers | 0.96 | 9,563 | 307 |
| no numbers | 0.90 | 11,936 | 288 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| a percentage or a money figure | 1.05 | 7,357 | 335 |
| no percentage or money figure | 0.99 | 32,501 | 316 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| 4 or more list lines | 1.10 | 9,922 | 352 |
| no list lines | 0.97 | 26,727 | 310 |
| 1 to 3 list lines | 0.92 | 3,209 | 293 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| carries a quoted passage | 1.06 | 10,684 | 338 |
| no quoted passage | 0.98 | 29,174 | 313 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| you 4 or more per 100 words | 1.08 | 13,538 | 344 |
| you 2 to 4 per 100 words | 1.06 | 10,211 | 337 |
| you under 2 per 100 words | 1.00 | 10,310 | 320 |
| never says you | 0.72 | 5,799 | 229 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| I under 2 per 100 words | 1.10 | 13,186 | 352 |
| I 4 or more per 100 words | 1.04 | 9,319 | 333 |
| I 2 to 4 per 100 words | 1.03 | 7,141 | 329 |
| never says I | 0.82 | 10,212 | 261 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| 4 or more question marks | 1.06 | 4,034 | 339 |
| one question mark | 1.05 | 10,184 | 335 |
| 2 or 3 question marks | 1.03 | 7,882 | 328 |
| no question marks | 0.94 | 17,758 | 300 |
Specific beats general: digits, a figure with a unit, a list the eye can count. Address the reader more than yourself. Question marks look like a lever on the raw index and vanish within the author (section 5).
Reposts and quote posts
An original post indexes 1.06; a repost 0.83; a quote post (a repost with the author's own words above it) 0.20, a median of 64 engagements against 339 for an original. Reposts are 1.3% of the classified posts.
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| regular | 1.06 | 32,331 | 339 |
| repost | 0.83 | 6,760 | 265 |
| quote | 0.20 | 767 | 64 |
A repost is a fifth of an original and a quote post a twentieth. If something is worth sharing it is worth writing about in a post of your own.
Timing: hour, day and slot
12 to 17 UTC indexes 1.16 against 0.87 for every other hour, and by the hour the peak is 14:00 UTC at 1.20 with the trough at 01:00 UTC (0.53). Evening (17 to 22 UTC) reads 0.76 and night 0.75. By day, Monday leads at 1.06, Sunday 1.05 and Friday 1.04; Thursday trails at 0.93. The strongest single cell is Sunday 12 to 17 UTC at 1.34, then Saturday 12 to 17 UTC at 1.24; the weakest is Thursday evening at 0.68.
Two checks temper it. Within the author, afternoon posts earn 1.07 of the same author's posts at other hours (84 of 141 authors do better in the afternoon), so the lever is real but small. Per follower, the afternoon reads 0.93: the biggest accounts also post then, which inflates the raw index. Timing is advisory in NBK's own rules for exactly this reason.
| Hour | Index | Posts | Median engagement |
|---|---|---|---|
| 00:00 UTC | 0.67 | 362 | 215 |
| 01:00 UTC | 0.53 | 321 | 170 |
| 02:00 UTC | 0.76 | 308 | 244 |
| 03:00 UTC | 0.78 | 379 | 249 |
| 04:00 UTC | 0.89 | 951 | 284 |
| 05:00 UTC | 0.92 | 1,687 | 295 |
| 06:00 UTC | 0.95 | 2,402 | 303 |
| 07:00 UTC | 1.03 | 1,825 | 329 |
| 08:00 UTC | 0.97 | 1,570 | 310 |
| 09:00 UTC | 1.06 | 1,545 | 337 |
| 10:00 UTC | 0.85 | 1,573 | 270 |
| 11:00 UTC | 0.96 | 1,855 | 306 |
| 12:00 UTC | 1.12 | 2,546 | 357 |
| 13:00 UTC | 1.16 | 4,289 | 370 |
| 14:00 UTC | 1.20 | 5,396 | 382 |
| 15:00 UTC | 1.19 | 4,135 | 381 |
| 16:00 UTC | 1.05 | 2,284 | 336 |
| 17:00 UTC | 0.77 | 1,659 | 246 |
| 18:00 UTC | 0.77 | 1,419 | 245 |
| 19:00 UTC | 0.70 | 1,038 | 223 |
| 20:00 UTC | 0.82 | 787 | 261 |
| 21:00 UTC | 0.77 | 590 | 246 |
| 22:00 UTC | 0.72 | 489 | 229 |
| 23:00 UTC | 0.70 | 448 | 222 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| Mon | 1.06 | 6,494 | 337 |
| Sun | 1.05 | 3,194 | 336 |
| Fri | 1.04 | 5,984 | 331 |
| Sat | 1.02 | 3,810 | 324 |
| Tue | 0.99 | 6,477 | 317 |
| Wed | 0.98 | 7,016 | 312 |
| Thu | 0.93 | 6,883 | 298 |
| Day | 05 to 08 UTC | 08 to 12 UTC | 12 to 17 UTC | 17 to 22 UTC | 22 to 05 UTC |
|---|---|---|---|---|---|
| Mon | 1.02 | 1.10 | 1.16 | 0.81 | 0.82 |
| Tue | 0.98 | 0.89 | 1.15 | 0.76 | 0.80 |
| Wed | 0.97 | 0.88 | 1.12 | 0.73 | 0.79 |
| Thu | 0.92 | 0.87 | 1.11 | 0.68 | 0.68 |
| Fri | 1.00 | 1.05 | 1.16 | 0.76 | 0.72 |
| Sat | 0.90 | 0.94 | 1.24 | 0.79 | 0.73 |
| Sun | 0.80 | 0.92 | 1.34 | 0.89 | 0.71 |
Each cell is the band’s median engagement against the learning set’s median. UTC throughout; the authors post from many time zones.
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| 12 to 17 UTC | 1.16 | 6.7 | 0.93 | 20.4% | 1.09 | 18,640 |
| other hours | 0.87 | 7.8 | 1.08 | 17.0% | 0.91 | 21,189 |
| Band | Index (raw) | Per 1,000 followers | Index per follower | Comments share | Comments index | Posts |
|---|---|---|---|---|---|---|
| Mon | 1.05 | 8.1 | 1.12 | 20.0% | 1.07 | 6,491 |
| Sun | 1.05 | 6.1 | 0.84 | 16.5% | 0.88 | 3,194 |
| Fri | 1.03 | 7.1 | 0.98 | 18.1% | 0.97 | 5,981 |
| Sat | 1.01 | 6.9 | 0.95 | 16.5% | 0.88 | 3,807 |
| Tue | 0.99 | 7.3 | 1.01 | 20.2% | 1.08 | 6,473 |
| Wed | 0.98 | 7.3 | 1.01 | 18.8% | 1.00 | 7,008 |
| Thu | 0.93 | 7.1 | 0.98 | 19.7% | 1.05 | 6,875 |
Early afternoon UTC, any day but Thursday, and the weekend afternoon slots read highest of all. Then stop optimising: the same author moves seven points by timing and the rest of this study moves more.
Cadence and the gap between posts
Authors who post six to seven times a week have the highest median of medians (379, 57 authors); four to five a week 258; eight or more 283; under one a week 139. The forty strongest authors by median post 6.2 times a week. Within the author, posts in a busier-than-usual week earn 0.93 of posts in a quieter week (48 of 131 authors do better when busier): cadence is a property of authors, and posting more in a given week does not lift the posts in it.
The gap matters more than the count. A post 20 to 30 hours after the author's last indexes 1.13; 12 to 20 hours 0.97; under six hours 0.80; two days 0.98; four to eight days 0.93; more than eight days 0.63. Within the author, a post under twelve hours after the previous one earns 0.82 of their posts twenty or more hours apart (31 of 131 authors do better fast).
| Band | Authors | Typical posts a week | Median of author medians |
|---|---|---|---|
| under 1 a week | 18 | 0.4 | 139 |
| 1 to 2 a week | 9 | 1.5 | 233 |
| 2 to 3 a week | 22 | 2.8 | 228 |
| 4 to 5 a week | 33 | 4.4 | 258 |
| 6 to 7 a week | 57 | 6.4 | 379 |
| 8 or more a week | 37 | 13.6 | 283 |
| Band | Index | Posts | Median engagement |
|---|---|---|---|
| under 6 hours since the last post | 0.80 | 6,601 | 255 |
| 6 to 12 hours | 0.92 | 2,203 | 293 |
| 12 to 20 hours | 0.97 | 2,947 | 310 |
| 20 to 30 hours (the next day) | 1.13 | 17,655 | 362 |
| 30 to 54 hours (two days) | 0.98 | 4,653 | 314 |
| 2 to 4 days | 0.98 | 2,843 | 312 |
| 4 to 8 days | 0.93 | 1,835 | 298 |
| over 8 days | 0.63 | 934 | 200 |
| first post in the sample | 0.74 | 187 | 236 |
One post a day, the next day, is the strongest rhythm in the set; twice in a day costs the second post a fifth. A week away costs more than that. Cadence is the habit of the strongest authors rather than a switch any one post can flip.
Topics
Each post was tagged for nine subject families by rule. Life outside work indexes 1.15 raw and 1.11 within the author, the strongest subject on both; failure and lessons 1.13; AI 1.05; leadership and teams 1.03 (37% of posts, the most common subject); hiring and careers 1.03; money 1.01. Content craft reads 0.90, LinkedIn itself 0.89 and agency and client life 0.87: the feed's own trade talk is its weakest subject.
11% of the learning set (4,240 of 39,858 posts) is time-bound, carrying a phrase such as "today", "this week" or "just launched"; the rest is evergreen.
| Subject | Share of posts | Index (raw) | Index per follower | Within the author | Comments index |
|---|---|---|---|---|---|
| life outside work | 16% | 1.15 | 1.03 | 1.11 | 0.94 |
| failure and lessons | 21% | 1.13 | 1.09 | 1.03 | 1.16 |
| AI | 21% | 1.05 | 0.71 | 0.99 | 0.92 |
| leadership and teams | 37% | 1.03 | 0.99 | 1.01 | 1.12 |
| hiring and careers | 31% | 1.03 | 1.01 | 1.04 | 1.11 |
| money | 27% | 1.01 | 0.96 | 1.00 | 1.14 |
| a platform by name | 5.8% | 1.00 | 0.88 | 1.01 | 0.84 |
| content craft | 30% | 0.90 | 1.06 | 1.00 | 1.26 |
| LinkedIn itself | 19% | 0.89 | 1.19 | 1.00 | 1.45 |
| agency and client life | 18% | 0.87 | 1.15 | 1.00 | 1.27 |
| Subject | Posts |
|---|---|
| leadership | 13,589 |
| hiring and careers | 11,329 |
| content craft | 11,195 |
| money | 9,848 |
| ai | 8,066 |
| linkedin itself | 7,816 |
| failure and lessons | 7,464 |
| agency and client | 6,495 |
| life outside work | 5,941 |
| platform news | 2,034 |
The subjects that win are the ones that are not about the work: a life outside it, a thing that went wrong. Posts about content, about LinkedIn and about agency life read lowest, which is worth knowing for anyone whose job is all three.
The profiles
The 44 strongest authors by median engagement were compared with all 344 profiles on 26 signals. Verification: 89% against 76%. Premium: 64% against 53%. A Top Voice badge: 16% against 5%. Creator hashtags set: 73% against 44%. A website link: 23% against 13%. Creator mode as recorded at harvest: 82% against 56%.
The headline is shorter: a median of 101 characters against 123. "I help" headlines are rarer (11% against 20%), a number in the headline a little more common (41% against 37%), and pipes or separators less common (59% against 70%). The About is longer, 1,402 characters against 1,208, with a shorter first line (92 characters against 102), and 73% carry a number in it. Median followers are 104,488 against 13,028, which is the caveat on every row: these are the profiles of accounts that are already large.
| Signal | Strongest | All profiles |
|---|---|---|
| Creator mode recorded on the profile | 82% | 56% |
| Verified | 89% | 76% |
| Premium | 64% | 53% |
| Top Voice badge | 16% | 5% |
| Creator hashtags set | 73% | 44% |
| Website link set | 23% | 13% |
| Education listed | 89% | 91% |
| Certifications listed | 39% | 47% |
| Has a banner | 95% | 93% |
| Headline length (characters) | 101 | 123 |
| About length (characters) | 1,402 | 1,208 |
| Positions listed | 5 | 5 |
| Connections | 10,935 | 5,622 |
| Followers | 104,488 | 13,028 |
| Creator hashtags (count) | 5 | 0 |
| Top skills shown (count) | 0 | 3 |
| Education entries (count) | 2 | 2 |
| Experience description (characters) | 345 | 398 |
| "I help" headline | 11% | 20% |
| A number in the headline | 41% | 37% |
| Pipes or separators in the headline | 59% | 70% |
| Role, company and a proof point in the headline | 9% | 8% |
| About first line (characters) | 92 | 102 |
| A list in the About | 34% | 31% |
| A call to action in the About | 25% | 25% |
| A number in the About | 73% | 66% |
Shares are the percentage of profiles with the signal; medians are the median value. Creator mode is reported as the harvest recorded it on the profile.
| Skill | Authors |
|---|---|
| Business Strategy | 4 |
| Leadership | 4 |
| Business Development | 3 |
| Entrepreneurship | 3 |
| Personal Branding | 3 |
| Communication | 2 |
| Digital Marketing | 2 |
| Innovation Development | 2 |
| Project Management | 2 |
| Social Media Marketing | 2 |
| Startups | 2 |
| Strategic Partnerships | 2 |
| ACATS | 1 |
| Artificial Intelligence (AI) | 1 |
| Brand Development | 1 |
| Hashtag | Authors |
|---|---|
| #entrepreneurship | 10 |
| #business | 9 |
| #marketing | 6 |
| #personalbranding | 6 |
| #socialmedia | 4 |
| #ai | 3 |
| #branding | 3 |
| #copywriting | 3 |
| #digitalmarketing | 3 |
| #growth | 3 |
| Measure | Strongest | All |
|---|---|---|
| Has a banner | 42 of 44 | 321 of 344 |
| Headline length, median characters | 101 | 123 |
| About length, median characters | 1,402 | 1,322 |
| Piped headline | 55% | 68% |
| Positions listed, median | 5 | 5 |
A profile cannot be A/B tested against itself, so every row here is prevalence: what the strongest accounts have, not what made them strong. The pattern is still plain. Verified, a short headline that states what you do without "I help", a long About that opens short and carries a number, and a banner.
The strongest authors against the typical author
Set the 40 strongest authors beside the 136 typical ones and most style habits are shared: media on 97% and 97% of posts, emoji on 66% and 62%, 12 and 11 blank lines, 6.3 and 5.8 posts a week. The differences are the levers this study found elsewhere: hashtags on 6% of posts against 10%, a first line of 47 characters against 54, a question close on 2% against 4%, and 1,162 characters against 1,090. Median followers: 114,195 against 27,650.
| Measure | Strongest 40 | Typical |
|---|---|---|
| Authors | 40 | 136 |
| Median followers | 114,195 | 27,650 |
| Median characters a post | 1,162 | 1,090 |
| Median first line, characters | 47 | 54 |
| Median blank lines | 12 | 11 |
| Posts a week | 6.3 | 5.8 |
| Posts with media | 97% | 97% |
| Posts with emoji | 66% | 62% |
| Posts with hashtags | 6% | 10% |
| Posts ending on a question | 2% | 4% |
The strongest authors do not write differently in most measurable ways. They hashtag less, open shorter, close on a statement and write a little longer. Everything else is shared, which is why the first-line and format sections carry the weight of this study.
Do the two cohorts agree?
The learning set is two groups of authors: a craft cohort (115 authors, 24,649 posts, median engagement 345) chosen for how they write, and a domain cohort (72 authors, 15,209 posts, median 288) chosen for working in social and marketing. A finding that holds in both is a finding about the feed rather than about one crowd. The levers this study leans on hold in both.
| Band | All | Craft | Domain |
|---|---|---|---|
| no hashtags | 1.05 | 1.05 | 1.05 |
| 1 to 3 hashtags | 0.82 | 0.78 | 0.88 |
| first line under 40 chars | 1.13 | 1.11 | 1.10 |
| first line over 140 | 0.76 | 0.77 | 0.76 |
| short lines (under 60 chars each) | 1.08 | 1.06 | 1.05 |
| paragraphs (over 110 chars a line) | 0.85 | 0.88 | 0.84 |
| over 8 blank lines | 1.11 | 1.10 | 1.10 |
| 1 to 3 blank lines | 0.66 | 0.63 | 0.70 |
| over 2000 | 1.29 | 1.31 | 1.20 |
| under 300 chars | 0.59 | 0.53 | 0.67 |
| has an image | 1.03 | 1.03 | 1.01 |
| text only | 0.71 | 0.59 | 0.94 |
| carries a link | 1.18 | 1.32 | 1.05 |
| quote | 0.20 | 0.20 | 0.19 |
| afternoon, 12 to 17 UTC | 1.16 | 1.20 | 1.08 |
| Thu | 0.93 | 0.90 | 0.97 |
Every band above moves the same way in both cohorts. The size of the move differs by a few hundredths, the direction never does.
Frequently asked questions
What does the study measure?
The public engagement (reactions plus comments) of 74,119 LinkedIn posts from 346 prominent writers, collected on 10 September 2026, grouped into bands by how a post is written, formatted and timed, and by what its author's profile carries. Each band's median engagement is indexed against the learning set's median (320), and checked per 1,000 followers, on the comments share and within the author.
Does the study name the authors or quote the posts?
No. The study publishes aggregate tables only. No author, handle, post, first line or follower count is published, and the export that builds this page refuses to write a file carrying one.
Is a band index a cause?
No. Every index is correlational across accounts of very different sizes. The within-author lens, each post against its own author’s median, is the honest check and is shown wherever it was computed; where the two disagree, the within-author reading is the one to trust.
Do hashtags help on LinkedIn?
No. Posts with no hashtags index 1.05; one to three index 0.82 and more than three 0.84, and the comment share falls to 0.71. Both cohorts agree.
Should the link go in the comments?
It makes no measurable difference (0.97). What costs engagement is the link card: 0.52 of the same author's image posts. A URL in the last fifth of the text of an image post is the least costly way to carry one.
How long should a LinkedIn post be?
700 to 2,000 characters leads per follower (1.12). Over 2,000 characters reads highest raw (1.29) and 0.88 per follower, a habit of the largest accounts. Under 300 characters indexes 0.59.
When is the best time to post on LinkedIn?
12 to 17 UTC indexes 1.16, with 14:00 UTC the peak hour and Sunday afternoon the strongest cell. Within the same author the afternoon is worth 1.07, so timing is a small lever next to the first line, the format and the hashtags.
Who is behind the study?
What Works on LinkedIn is researched by Matt Cunnelly and Josh Stoddard and published by NBK Social Research, the research imprint of NBK Social, a social operations agency working between Dubai and Manchester.
NBK Social Research (2026). What Works on LinkedIn: 74,119 posts measured. nbksocial.com/research/what-works-on-linkedin/