A client sent us a screenshot last month. Amazon had announced a fee change at 10am. He published a sharp, correct post about it at 11:40am. It reached 1,100 members. His baseline is a little over 5,000.
His read was that he'd been too slow. Somebody else had got there first, the algorithm had rewarded them, and he'd missed the window.
That is not what happened. He was not slow. He was one of about forty people in his network who published a post that said the same thing on the same afternoon, and LinkedIn's feed is built specifically to stop a reader seeing all forty.
This is the LinkedIn diversity re-ranker, and for ecommerce founders it is the least discussed mechanic in the stack, because it only bites on the days you feel most compelled to post.
What LinkedIn has actually documented
We are careful about platform claims, so here is what is on the record and what is not.
LinkedIn's engineering blog describes the feed as a pipeline. First-pass rankers pull candidates. A second-pass ranker scores them for the individual reader. Then a final re-ranking layer modifies that list before it renders. In LinkedIn's own description of its ranking stack, that last layer is called the diversity re-ranker, and its stated job includes injecting diverse content into the top positions and reducing duplication of similar content.
That is the documented part. What is not documented is any weighting, any threshold for "similar," or any figure for how much reach it costs you. Nobody outside LinkedIn has that number. If someone quotes you a percentage for the same-topic penalty, they invented it.
What survives the caveats is the shape. The feed is not only asking "is this post good for this reader." It is also asking "has this reader already seen something like it." On a normal Tuesday the answer is no, because nothing in the candidate set looks like your post. On the day Amazon changes a fee, the answer is yes forty times over.
Why ecommerce founders hit this harder than anyone
Most people's networks do not share a calendar. Yours does.
Amazon publishes a fee schedule. Every founder in your lane gets the same Seller Central notification at the same minute. Prime Day dates land in one press release. A policy changes on a dated page. Earnings drop at 4pm on a Thursday. The raw material is identical, the timing is identical, and the people you are connected to are precisely the people who also received it.
So on those days your post enters a candidate set that is unusually full of things that look like it. The re-ranker is doing its job, which is protecting the reader from eight consecutive takes on the same announcement. The cost of that job lands on the seven takes that did not get the slot.
Notice what this does to the instinct every founder has on a news day. The urge to post is highest exactly when the feed is most actively deduplicating. The feeling that you must say something about this today is a signal that everyone else has the same feeling, and the feed will show each reader one of you.
What the re-ranker keeps and what it folds
We have watched this across enough client accounts to say which posts survive a crowded day and which get folded away. The pattern is consistent, and it is not about speed.
What gets folded:
- The restatement. "Amazon just announced X. Here's what it means." The reader has already been shown this. Yours is a duplicate of a duplicate.
- The hot take. Positions cluster too. On fee-change day there are two available opinions and twenty people holding each. A position is one of the most copyable things you can publish, and on a news day it is being copied in real time.
- The framework post. "Three things sellers should do now." Every agency in your feed published one by 2pm, and they all say pull your margins, check your inventory, talk to your accountant.
What survives:
- The derivation. What this change costs on a specific catalogue, with the arithmetic shown. "On our 40-SKU line this moves contribution by $0.31 a unit on the top three and nothing on the rest, because the change only reaches the size tier those three sit in." Nobody else has your catalogue. The post cannot be similar to anything.
- The boundary. Who this does not apply to. Forty posts say it matters; the one that says "if you're under 1 lb this is a non-event and here's why" is structurally different from all of them.
- The dated contrast. "This is the third change to this fee since 2024. Here's what the previous two did to our number." That requires you to have been there, and being there is not something the re-ranker can find elsewhere.
All three come from the same place: your access. We have written before that access is the only layer of a claim nobody can restate. This is the mechanical reason it matters. Similarity is measured on content, and the only content that cannot be similar to a stranger's is the content that came out of your business.
The second-day post is the underpriced slot
Here is the move most founders never make.
On news day the candidate set is packed and every reader's feed has already been deduplicated down to one or two takes. Three days later the candidate set is empty. Nobody is writing about the fee change anymore because it is old news. And the readers who saw the announcement on Tuesday have now had 72 hours to work out that they do not know what it means for them.
A post on day three that says "I've now read the actual fee page, run it against our catalogue, and here is the part every summary got wrong" lands in a feed with no competitors and in front of a reader who is more interested than they were on Tuesday, not less.
This is not "be slow." It is a two-post rule. On the day, only publish if you have a number nobody else can have. If you do not, hold it. On day two or three, publish the derivation. The founders we work with who run this consistently see the day-three post outperform the day-one post on members reached, and it is not close.
We wrote up a newsjacking system some time ago, and it still holds. What this mechanic adds is a filter on what the newsjack has to contain. Speed gets you into the candidate set. Access is what gets you the slot.
What this mechanic does not do
Three things, so nobody over-corrects.
It is not a penalty. Nothing is flagged, nothing carries forward. A folded post on Tuesday does not depress Thursday. That is a different mechanic, and it is not this one.
It does not apply on normal days. If your post is about a return-reason pattern you found on a Wednesday in a category nobody else in your network is in, there is nothing in the candidate set for it to be similar to. Most of your calendar is unaffected. This is a news-day problem.
It does not hide you from people who already follow you closely. First-degree connections with a history of engaging with you are the most likely to be shown your version. The cost is in the beyond-network distribution, which is the part news posts are usually published to win.
A five-minute check on your own account
Open your analytics. Find your last three posts that were pegged to something everyone in your lane also wrote about the same day. Note members reached on each. Now compare them to the three posts before and after that were not news-pegged.
For most founders the news-pegged posts sit below baseline, sometimes well below, and the founder has filed each one under "the algorithm was weird that week." It was not weird. It was doing exactly what it is documented to do.
Then read the three news posts and ask one question: could forty other people have written this from the same notification? If yes, you now know why.
FAQ
Is this the same thing as the off-lane cost? No. The off-lane cost is what happens when you leave your subject. This is what happens when everyone arrives in it at once. Opposite conditions, and they have opposite fixes. Off-lane is corrected by returning to the subject. This is corrected by bringing something to the subject that nobody else can.
Should I stop posting about Amazon news? No. Stop posting the announcement. The news is the peg, not the post. If the only thing you can add is that it happened, you are competing with the notification itself, and the notification got there first.
Does mentioning my brand or my category make the post different enough? Not in the way you hope. Similarity is about what the post says, not whose name is on it. "Here's what this means for supplement sellers" is still a restatement if the body is the same summary with a category word swapped in.
What about reposting someone else's post with my thoughts? On a news day, a repost with commentary is a duplicate by construction. You have attached your name to content that is already in the candidate set. Write the derivation as your own post instead, or wait until day three.
How do I know if the day-three post is working? Members reached, not impressions. A day-three post that travels will show a higher share of reach from outside your network than the day-one version did. That is the re-ranker having nothing to fold it against.
We build LinkedIn content systems for ecommerce founders, and a running order that knows which days are crowded is part of what we build. If your news-day posts keep underperforming and you have been blaming timing, the fix is usually not speed. It is what the post contains. If you want a second opinion on your last quarter of posts, we are happy to look.