The best time to post on LinkedIn, from 1,839 public posts.

• Weekday timing is close to flat. Monday to Friday, the median post earned between 27 and 34 reactions and comments. The gap is smaller than the noise.
• Weekends ran higher, on far fewer posts. Saturday's median was 53 on 141 posts and Sunday's 69 on 129, against 34 on a weekday. The two public studies below say the opposite for their populations. Read on before you move to Sundays.
• Early UTC did better than the UTC afternoon. The 00:00 to 02:59 band had a median of 71; the 21:00 to 23:59 band had 20. Early UTC is the European and Gulf morning, and where these 95 people live we do not know.
• The hour moves the number less than the author does. Between the best and worst three-hour band the median gap is 51. Between a typical author and one in the top tenth it is 176. And inside one author's own history, the "good" hours lost more often than they won.

So: post in the morning of the people you want to reach, hold that slot for six weeks, and then read your own numbers, which are the only ones that describe your audience. The rest of this page shows the work.

Sonar keeps a radar of public LinkedIn posts for every founder who uses it: the people on their watchlists, a weekly snapshot of the top posts in their niche, and a shared library of high-performing posts on founder topics. On 22 September 2026 we read those public captures and nothing else. No founder's own posts, no private analytics, nothing tied to a workspace or a person.

• Sources. 1,861 watchlist captures, 349 niche-snapshot posts and 918 library posts, deduplicated by URL to 3,104 unique posts published between 31 December 2025 and 22 September 2026. Reposts were dropped, since a repost's timestamp belongs to the resharer.
• Why the tables use 1,839 of them. The library keeps only posts above an engagement floor, so its median is 1,416.5 against the watchlist's 34, and the niche snapshot sits at 12, a third population again. Pool them and a bucket's median tracks how many library posts landed in it, which says nothing about the hour. The watchlist capture is everything a watched person posted in the window, with no filter on how well it did, so every timing table below reads that set: 1,839 posts by 95 people, 20 March 2026 to 21 September 2026.
• UTC. A public post carries the moment it was published, and no timezone for the person who published it. Every hour on this page is UTC. The conversions further down are a reading aid, and only that.
• Medians. Engagement is reactions plus comments. The distribution is long-tailed: the mean of the whole set is 146 and the median is 34, because a handful of posts earned thousands. A median describes the typical post; a mean describes the outliers. Both are printed so you can see where they part.
• Thin buckets. Any bucket under 30 posts would be marked thin and printed fainter. None of the weekday or three-hour buckets below is under 30; the single-hour buckets in the underlying data include some that are, and we drew no conclusions from them.

Sorted Monday to Sunday. The weekday medians sit within a few points of each other. The weekend rows are the ones to be careful with: higher medians, less than half the volume.

Two readings fit the weekend rows and the data cannot choose between them. One: fewer posts compete for the feed on a Sunday, so the ones that go out get more of it. Two: the kind of person who posts on a Sunday is a different kind of poster, and would earn more on a Tuesday as well. The within-author test further down leans toward the second.

Single hours are too thin to trust on their own, so the day is cut into eight bands. The third column is every day; the fourth is Monday to Friday only, for the reader who has already decided against weekends.

The shape is one slope: the bands before 09:00 UTC earned more than the bands after 12:00 UTC, and the Monday to Friday column keeps the shape. The mean column shows where a single post distorts a band; 15:00 to 17:59 has a mean of 193.7 on a median of 24, which is one or two enormous posts and a great many ordinary ones.

Before you act on the slope: early UTC is the working morning for Europe, the Gulf and India, and the middle of the night for the Americas. Sonar's founders and the people they watch skew toward the first group. If your audience is in New York, a band that reads "best" here is 02:00 for them, and this table is telling you about someone else's morning.

Every best-time table, this one included, is a pile of different people posting at different hours. The hour and the person arrive together, and a table cannot pull them apart. Two more cuts of the same data show how much of the pattern is the person.

The author spread dwarfs the hour spread

Take the 62 people with five or more posts and find each one's own median. A quarter of them sit at 6 or below. The middle author sits at 24. The top quarter starts at 66 and the top tenth at 200. That is a range of 194 at the median between people, against a range of 51 between the best and worst three-hour band. Who is writing, to whom, about what, moves the number by an order of magnitude more than when they pressed Post.

Inside one person's history, the good hours mostly lost

The fairest test asks each person about their own posts. For everyone with at least 20 posts and at least five in each window, we compared their median before 09:00 UTC (the winning bands above) with their median between 12:00 and 20:59 UTC (the losing ones). The early window won for 6 people, the later window won for 10, and one tied. The aggregate slope did not hold up for the majority of the individuals inside it. 17 people is a small test, and we would run it again on more; it is enough to stop anyone treating the band table as an instruction.

Posting more often went with a slightly higher typical post

Bucket the same 62 people by how often they posted across their own span. Under once a week (26 people), the median author's median was 22. One to three a week (22), 22.75. Three or more a week (14), 27.5. A direction on small groups, and the direction every practitioner would predict: the people who post most are also the people whose typical post does best. Frequency did not cost them.

Two of the large scheduling platforms publish their own best-time studies. They measure their own customers' posts, which is a different population from ours, and they do not agree with each other on the hour.

• Sprout Social (published 31 March 2026) says the best time to post on LinkedIn is "Tuesdays from 11 a.m. to 5 p.m.", names Tuesday, Wednesday and Thursday as the best days and weekends as the worst, and drew on "nearly 2 billion engagements across 307,000 social profiles" across six networks between 27 November 2025 and 27 February 2026. Sprout Social's best times to post on LinkedIn.
• Buffer (published 9 September 2026) puts the peak between 3 p.m. and 8 p.m. on weekdays in the poster's local time, with Wednesday at 4 p.m. the strongest slot and Monday the weakest day, from "over 4.8 million posts sent through Buffer". Buffer's best time to post on LinkedIn.

Late morning against late afternoon, on millions of posts each. Both are probably right about the people they measured, which is the point. Both agree that weekends are weaker for their customers, where our unfiltered watchlist read the other way on a much smaller sample. A best time is a property of an audience. Yours has one, and it lives in your own analytics.

1. 01

   Name the audience, then their morning

   Decide who you want reading: buyers in one country, peers across a region. Take the hour they open LinkedIn with coffee, 7 to 9 in their own time. If they span continents, pick the larger group and stop trying to serve both with one slot.

2. 02

   Hold one slot for six weeks

   Post at that hour, three or more times a week, and change nothing else about the timing. A best time measured on four posts is noise; on eighteen it starts to be yours.

3. 03

   Read impressions, then engagement

   Open LinkedIn's post analytics each week and note impressions in the first 24 hours, then reactions and comments after a week. Impressions tell you whether the slot reached people; engagement tells you whether the post deserved it.

4. 04

   Compare against your own median

   Your median post is the benchmark. A post at 1.5 times your median in a new slot is a signal; the same post beating a number from a stranger's table tells you nothing.

5. 05

   Move one slot at a time

   If the morning slot underperforms your median for six weeks, try the audience's lunch hour for the next six. One variable, one block of weeks, one honest comparison.

Two of the steps need a number LinkedIn does not show on the post itself. Impressions live in your own analytics, which you can read by hand each week or have read for you; what LinkedIn impressions mean covers how LinkedIn counts them. Your posting history and its engagement ratio are the other half. When you have both, the tables on this page stop mattering.

A draft saved in Sonar takes a default slot of 9:00 in your own timezone, which the app learns from your browser, so "the morning" means your morning and never the server's. Move the slot to whatever your audience's morning is, mark the post Scheduled, and a check every 15 minutes publishes it to LinkedIn once its time arrives. The LinkedIn platform page has the connection and the publishing in full.

Once six of your own dated posts are on file, the weekly Insights edition runs the same day-and-time cut on your history that this page ran on strangers, and prints it as one line: your posts land best on these days, in this block, from this many posts. The Sonar for LinkedIn extension adds the impressions LinkedIn shows only you. That line is the answer this page cannot give.