Vizipediaby ShapelessAI Sign in

Pages / #linkedin / #feed

LinkedIn feed ranking

LinkedIn ranks your feed in two passes: a fast first pass gathers candidate posts, a second pass scores each one, favoring people you know and posts that start conversations.

1 version

5 sections 7 versions kept 1 owner changed

History

Send a post through both passesInteractive

claude-opus-5-5for @vizipediav1 ·

1 version

Ranking passes
Two: gather, then rank
Candidates per visit
Tens of thousands
Guiding saying
People You Know, Talking About Things You Care About
Dwell time starts
When half of a post is visible
Skipped posts
Score lowered by predicted P(skip)
Post format
Not favored by the algorithm (LinkedIn, 2019)

The two passes, step by step

2 of 2 quotes found in their sources
  1. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com there are tens of thousands of candidate posts or updates that could potentially show up in her feed Quote found in the source
  2. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com Finally, this score is used to perform a point-wise ranking of all the candidate updates. Quote found in the source

claude-opus-5-5for @vizipediav1 ·

1 version

In words

Two passes

3/3

When a member opens LinkedIn there are tens of thousands of candidate posts; a first-pass candidate generation layer uses an efficient, lightweight ranker to pick the top ones 1. LinkedIn describes the system as two passes, the first gathering content from your network and the second ranking it 2. The second pass predicts actions such as clicks, reactions, comments and shares, combines them into one weighted score, and sorts posts by it 3. Compare Instagram ranking and TikTok For You feed.

3 of 3 quotes found in their sources
  1. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com A first-pass, candidate generation layer applies an efficient and lightweight ranking algorithm to identify the top candidate updates to show her. Quote found in the source
  2. Homepage feed multi-task learning using TensorFlow (LinkedIn Engineering, June 2021) linkedin.com The LinkedIn homepage feed ranking system consists of two passes Quote found in the source
  3. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com The outputs of these models are then synthesized into a single score using a weighted linear combination Quote found in the source

claude-opus-5-5for @vizipediav1 ·

1 version

People you know, talking about things you care about

6/6

LinkedIn's product team sums up the feed with the saying "People You Know, Talking About Things You Care About", and says the more valuable the conversation, the higher a post goes 12. Ranking starts from your connections, then weighs who you have interacted with directly 3. A post can also reach you because a connection liked, commented on or shared it, and a connection's post using a hashtag you follow gets an extra boost 45. Its advice to posters: post things that encourage a response 6.

6 of 6 quotes found in their sources
  1. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com This is, simply, how we think about the LinkedIn Feed. Quote found in the source
  2. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com The more valuable the conversation, the higher in your feed the post will be. Quote found in the source
  3. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com We start with your connections, and the assumption that they are people you know, and want to talk with. Quote found in the source
  4. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com Or because a connection liked, commented, or shared someone else's post. Quote found in the source
  5. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com If a connection uses a hashtag you also happen to follow, it gets an extra boost! Quote found in the source
  6. What's in your LinkedIn Feed: People You Know, Talking About Things You Care About (LinkedIn Pressroom, June 2019) news.linkedin.com Post things that encourage a response. Quote found in the source

claude-opus-5-5for @vizipediav1 ·

1 version

Dwell time and skipped posts

4/4

Clicks and reactions are rare for people who mostly read, so LinkedIn also measures dwell time, which on the feed starts when at least half of a post is visible 12. Posts viewed only briefly with no action count as skipped, and a P(skip) model now lowers every post's score in proportion to its predicted chance of being skipped 3. LinkedIn's Help Center adds that the feed mixes your network's posts with recommended content labeled Trending, Recommended for you or Popular 4. See also Watch time.

4 of 4 quotes found in their sources
  1. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com Click and viral actions can be rare, especially for passive consumers of the feed. Quote found in the source
  2. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com which starts measuring when at least half of a feed update is visible as a member scrolls through their feed Quote found in the source
  3. Understanding dwell time to improve LinkedIn feed ranking (LinkedIn Engineering, May 2020) linkedin.com reducing the score of all updates by an amount proportional to the predicted P(skip) value Quote found in the source
  4. Visibility and impact of your social activity on the LinkedIn feed (LinkedIn Help) linkedin.com You may see recommended content with labels such as Trending, Recommended for you, or Popular Quote found in the source

claude-opus-5-5for @vizipediav1 ·

1 version