Stop measuring LinkedIn content performance with likes

The number founders most often lead their weekly content reports with is impressions. It is the wrong number. LinkedIn does not count impressions as unique viewers: one person scrolling past your post three times in a session generates three impressions. The signals that decide your next post's reach are missing from the default dashboard.

Likes, Comments, and Impressions Don't Predict LinkedIn Content Distribution

The metrics on a typical LinkedIn report do not predict distribution. Impressions, likes, and follower count are numbers teams can see, not inputs the algorithm uses to decide who sees your next post. They correlate with reach weakly, and impressions in particular can climb while the number of unique people you reach stays flat.

This is not a marginal problem. 71% of B2B marketers say they cannot effectively measure content performance on LinkedIn, per the Content Marketing Institute's 2025 B2B Content Marketing Report. The most common mistake we see is a stakeholder report that leads with impressions and follower growth, two numbers that move independently of whether the content reached new people or influenced pipeline.

The ground under these metrics has also shifted. Between 2024 and 2025, LinkedIn organic reach fell hard: views down roughly 47%, engagement down 39%, and follower growth down 42%, per Richard van der Blom's analysis of 1.8 million posts. That is not a seasonal dip. It is a structural realignment toward interest-graph signals, where topic relevance decides distribution instead of how many connections you have.

There is a second, quieter cost to optimizing for likes. Public reactions reward content that seeks validation, not content that travels. The posts that pull the most likes are frequently not the posts reaching the most new people. When you tune your calendar for applause, you train yourself to write the wrong posts well.

What Is LinkedIn's Depth Score and How Does It Affect Organic Reach?

Depth Score is practitioner shorthand, not a term in LinkedIn's public documentation. It describes the composite quality signal LinkedIn's recommendation engine builds from time-weighted engagement behavior rather than raw counts. To measure it, you first need to know what is doing the measuring.

That engine is 360Brew, a 150-billion-parameter, decoder-only foundation model published by LinkedIn's own research team in January 2025. It handles 30 or more ranking and recommendation tasks at once, feed distribution among them, replacing the fragmented per-task machine learning systems that came before. One model now decides a large share of what surfaces in the feed.

LinkedIn evaluates each post in three sequential phases. First, quality classification in the minutes after posting. Second, an early engagement signal assessment in the first 30 to 60 minutes. Third, Depth Score accumulation over the following 24 to 48 hours, which either expands or contracts final distribution. The first two phases decide whether a post lives; the third decides how far it travels.

The practical takeaway is a reordering of what counts. A post with a few dozen saves and one deep comment thread accumulates more Depth Score than a post with hundreds of likes and a wall of one-line comments. 360Brew measures attention depth, not social proof volume. Once you internalize that, most of the standard advice about chasing reactions falls apart.

The Four Signals Behind LinkedIn Content Performance Metrics 2026

Four signals carry the weight: dwell time, saves, comment depth, and private shares. None of the four is the metric most teams report, and two of them are invisible in LinkedIn's native analytics.

Dwell time is the primary one. Posts where readers spend 61 or more seconds average a 15.6% engagement rate, against 1.2% for posts under 3 seconds, a 13 times difference. LinkedIn does not expose raw dwell time anywhere, but a high engagement-to-impression ratio on a document or long-form text post is a dependable proxy for it.

Saves carry roughly 5 times the algorithmic weight of a like and 2 times the weight of a comment, per Richard van der Blom's analysis of 1.8 million posts in the 2025 Algorithm Insights Report. LinkedIn has not published a weighting table, so treat these as practitioner inference, not platform disclosure. The direction is what matters: a save is a reader telling the algorithm the post was worth keeping.

Comment depth outranks comment count. LinkedIn evaluates threaded conversation substance and thread length, not the number of first-level comments. A post with one long reply thread outperforms a post with the same number of separate one-line reactions. Substance over tally.

Private shares, the Send to and Message actions, carry real distribution weight and are entirely invisible in native analytics at every tier, including the Analytics API. LinkedIn does not surface private share counts to creators at all. The only way to infer elevated private sharing is an anomalous reach-to-engagement ratio: a post drawing far more impressions than its like-and-comment rate would predict is almost certainly being forwarded in DMs. Track that ratio across a content series, not just per post, and the topic clusters that trigger private sharing become visible. Those clusters are your highest-distribution bets.

LinkedIn Impressions Are Structurally Misleading for B2B Reach Measurement

LinkedIn impressions are not unique views. One person scrolling past your post three times in a single day generates three impressions. That single fact makes the metric structurally misleading for reach assessment.

The consequence is that a high impression count can hide poor reach. A post can accumulate impressions by being shown repeatedly to a small, active slice of your followers rather than by entering new feeds. A smaller impression count paired with a strong engagement rate often reaches more first-time viewers than a larger impression count paired with a weak one.

LinkedIn's dashboard does not expose unique reach by default. The gap between reported impressions and actual unique viewers swings with how many of your followers are heavy platform users on a given day. On a busy day, your regulars inflate impressions without adding a single new reader.

Over a quarter, impression-forward reporting compounds the error. It overstates how far your content travels while hiding the fact that the same handful of audience segments are seeing the same posts again and again. For B2B teams trying to prove they reached a market, that is precisely the wrong picture.

Saves Over Likes: Why Your LinkedIn Content Strategy 2026 Needs a New Signal Hierarchy

If you rebuild your signal hierarchy around one metric, make it saves. They carry roughly 5 times the weight of a like, and they are the only depth signal you can directly incentivize without tripping LinkedIn's engagement bait detection.

First, a warning about what suppresses reach. Posts containing external links receive roughly 60% less reach. LinkedIn penalizes off-platform intent at the distribution layer, not only by lowering feed placement but by capping how much Depth Score the post can accumulate. If your posts routinely link out, that reduction compounds across your entire posting history.

Now the lever. LinkedIn's bait filter targets prompts that solicit public social proof, comment YES if you agree, react if this resonates. It does not penalize prompts that invite private reference behavior: save this checklist, screenshot the framework in slide 4. Ending a post with explicit, concrete utility language reliably lifts save counts 2 to 4 times without a distribution penalty.

The distinction is directional, and that is the part most people miss. Bait prompts pull engagement toward the public feed and toward other people performing approval. Utility prompts push value toward the individual reader who wants to come back to it. LinkedIn's classifier reads prompt direction, not prompt existence, so the honest ask, keep this, is safe while the hollow ask, prove you liked this, is not.

What the Typical LinkedIn Content Strategy 2026 Gets Wrong About Posting Time

Most LinkedIn posting advice treats time of day as a traffic question: when is the platform busiest? That framing quietly misses how distribution begins.

LinkedIn tests each post against an initial cohort of 2 to 5% of your network. That cohort's engagement pattern in the first hour decides whether the post advances to the next distribution tier. And the stakes are high: only 5% of posts that underperform in their first 60 minutes ever recover to reach a broader audience. The opening hour is close to deterministic.

Here is the part the traffic framing ignores. That initial cohort is not drawn randomly from your followers. It is heavily weighted toward your most consistent recent engagers. So posting time is not when is LinkedIn busiest, it is when are my specific seed engagers online. Post at peak general traffic but catch your power engagers stuck in a morning standup, and identical content underperforms.

Accounts that A/B test their posting windows against their top-10 engager activity patterns, rather than against generic best-time charts, see measurable distribution differences on otherwise identical posts. The chart tells you when the crowd is scrolling. It cannot tell you when the ten people whose early engagement decides your fate are scrolling.

Build a Depth Proxy Scorecard Using Native LinkedIn Analytics

You can build a usable depth scorecard from four things LinkedIn's native analytics already give you: engagement rate, comment count, save count where your plan tier exposes it, and the reach-to-impression ratio. Private shares stay invisible, but their signature is detectable, and so is comment depth.

For private shares, track each post's ratio of impressions to engagement. When a post pulls far more impressions than its like-and-comment rate predicts, it is almost certainly being forwarded in DMs. Log that ratio across a series and the topic clusters that trigger private forwarding surface on their own. Those are the topics quietly doing your best distribution work.

For comment depth, track your own reply rate and whether your replies generate second-level responses. Posts where you reply to 50% or more of comments, and where those replies pull further engagement, consistently earn a distribution bump 24 to 48 hours after publishing. That second-day reach spike is the operational signature of a high Depth Score post, corresponding to the accumulation phase.

A workable weekly scorecard is four columns: save count per post, save-to-impression ratio, reply-to-comment rate, and a yes or no for whether the post produced a second-day reach bump. Pulled straight from the native UI, those four give a cleaner signal picture than any impression-forward dashboard your analytics vendor will sell you.

Format Rankings After the March 2026 Authenticity Update

Document and carousel posts average a 6.60% engagement rate, the highest of any LinkedIn format. Native video follows at 5.60%. Text-only posts average 2.00%. Polls collapsed to 0.07% after LinkedIn's March 2026 Authenticity Update, down from previously competitive rates.

That update targeted content patterns that simulate engagement without generating it: polls, generic reaction prompts, and templated posts. A content calendar built around polling for audience insight now actively costs you reach. The format ranking tracks dwell time, documents make people scroll, video makes them watch, and a poll takes one tap and produces no meaningful dwell signal.

The reach penalty on AI content is more subtle than it looks, and it is not really about AI. It reads as pattern-matching on structural sameness. Posts using identical hook templates, predictable parallel bullet structures, and interchangeable transitions get down-weighted regardless of who or what wrote them. Fully AI-generated content can see roughly 2.8 times less reach. The likely mechanism: 360Brew's embedding space clusters structurally similar posts together, and an account that keeps producing content clustering with low-engagement patterns gets down-weighted proactively.

So the fix is not to avoid AI. It is to avoid sameness. Founders who write with distinctive vocabulary and unconventional structure, even with AI assistance, sidestep the penalty. What you are protecting is fingerprint diversity across your posting history, not the human origin of any single post.

Company Pages vs. Personal Profiles: The 13x Distribution Gap

Personal profiles receive roughly 65% of LinkedIn feed distribution. Company pages receive roughly 5%. That 13 times structural gap is the single biggest reason founder-led content outperforms brand pages on identical budgets.

The gap is not an oversight to be optimized away. LinkedIn is a professional and interest platform, not a brand broadcast channel. Its distribution model rewards people discussing ideas and largely declines to carry organizations promoting themselves. No amount of posting cadence closes a 13 times handicap.

For a B2B team deciding where to put content effort, this is close to decisive. A well-followed company page will typically lose to a founder's personal profile with a far smaller audience, because the personal post starts with a structural distribution advantage the page can never match.

The practical move for linkedin content strategy 2026 is to treat the company page as a credibility layer and content archive, not a distribution engine. Build reach through personal profiles, and link back to company page assets from there once the audience is already paying attention.

Frequently asked questions

What metrics predict LinkedIn content distribution in 2026?

The four metrics with the most predictive weight are dwell time, save rate, comment thread depth, and private shares. Dwell time over 60 seconds correlates with 15.6% engagement rates vs. 1.2% for posts under 3 seconds. Saves carry approximately 5x the algorithmic weight of a like. Impressions and follower count have no direct role in LinkedIn's distribution formula.

Does LinkedIn favor saves over likes for organic reach?

Yes. Practitioner analysis of 1.8 million posts (Richard van der Blom, 2025 Algorithm Insights Report) places saves at approximately 5x the algorithmic weight of a like and 2x the weight of a comment. LinkedIn has not published an official weighting table; these are practitioner-inferred ratios. Saves can also be directly incentivized through utility-framed post endings without triggering LinkedIn's engagement bait detection.

What is LinkedIn's Depth Score and how does it affect organic reach?

Depth Score is practitioner shorthand for the composite quality signal LinkedIn's 360Brew algorithm builds from time-weighted behaviors: dwell time, saves, comment thread depth, and private shares. LinkedIn does not use this term publicly. Posts with higher Depth Scores enter successive distribution tiers over 24 to 48 hours. Posts that underperform in their first 60 minutes face only a 5% chance of broader recovery.

Why are LinkedIn impressions a misleading metric for B2B content performance?

LinkedIn impressions are not unique views. One person scrolling past your post three times in a single session counts as three impressions. A high impression count can mask poor reach by reflecting repeated exposure to a small audience rather than broad distribution. For B2B reach assessment, engagement rate and the reach-to-impression ratio are more accurate indicators than raw impression volume.

How do I measure dwell time on my LinkedIn posts if LinkedIn does not show it?

LinkedIn does not expose raw dwell time to creators at any analytics tier. The practical proxy is engagement rate on long-form content: a high combined rate of likes, comments, and saves relative to impressions on a document post or text post over 300 words is a reliable indicator of extended dwell time. Tracking this ratio across a content series reveals which topics hold reader attention longest.

What is a good save rate benchmark for LinkedIn posts in 2026?

LinkedIn does not publish save rate benchmarks, and the metric is not universally surfaced in native analytics. As a rough practitioner reference: a save-to-impression ratio above 0.5% on text posts and above 1.5% on document posts suggests strong utility signaling. Posts ending with specific reference prompts, 'save this checklist,' 'screenshot slide 4,' tend to outperform posts with no save prompt by 2 to 4 times.

How does comment depth differ from comment count in LinkedIn's algorithm?

LinkedIn's algorithm evaluates threaded conversation substance and thread length, not just the total number of first-level comments. A post with one deeply threaded exchange, eight replies across three people, outperforms a post with eight separate one-line reactions. LinkedIn has not published documentation on this weighting, but the signal is consistent with 360Brew's attention-depth model. Tracking how often your replies generate second-level responses is the practical proxy.

Can I track private shares, the 'Send to' or 'Message' actions, in LinkedIn analytics?

No. LinkedIn does not surface private share counts to creators at any access level, including the LinkedIn Analytics API. The only way to infer elevated private sharing is by monitoring an anomalous reach-to-engagement ratio. When a post receives significantly more impressions than its like-and-comment rate would predict, private DM forwarding is the most likely explanation. Tracking this ratio across a content series identifies which topic clusters trigger private sharing.

How has LinkedIn organic reach changed between 2024 and 2026, and why?

LinkedIn organic reach declined sharply: views fell approximately 47%, engagement 39%, and follower growth 42% between 2024 and 2025, per Richard van der Blom's analysis of 1.8 million posts. The cause is structural: LinkedIn shifted its feed toward interest-graph signals over social-graph signals. Posts now reach people based on topic relevance rather than connection proximity. Content built for a social-graph model, posts targeting your immediate network, underperforms in an interest-graph model.

What content formats get the highest engagement on LinkedIn in 2026?

Document and carousel posts average 6.60% engagement, the highest of any format. Native video follows at 5.60%, and text-only posts average 2.00%. LinkedIn polls collapsed to 0.07% after the March 2026 Authenticity Update. The format hierarchy reflects LinkedIn's preference for content generating extended dwell time: documents require scrolling through slides, video requires sustained watching, while polls require only a single tap and generate no meaningful dwell signal.

Sources and further reading

• LinkedIn's 360Brew foundation model research (arXiv, January 2025)
• Content Marketing Institute 2025 B2B Content Marketing Report
• LinkedIn Creator Analytics: what's measured and how