LinkedIn Analytics: the 8 metrics that matter (and those I ignore)

LinkedIn's analytics dashboard shows forty metrics. Impressions, engagement rate, reach, demographic breakdown, reactions by type, comments, shares, link clicks, profile views, new followers per week, engagement by hashtag...

The problem: most of these metrics aren't useful for decision-making. They're vanity metrics that feel good but don't tell you what to do.

After 3 years of LinkedIn performance analysis on thousands of accounts, the metrics I watch weekly are 8. The rest I ignore. In this article I explain why.

1. The framework: input-output-outcome

Before choosing metrics you need a reading framework. The most useful one divides metrics into 3 levels:

Input (what you do)

Number of posts published, hours invested, tool cost. These metrics tell how much you're investing, not if it works. Useful only as context.

Output (what it produces)

Impressions, engagement, comments, shares. These tell how LinkedIn distributed and made your content react. They're proxies of value, not direct value.

Outcome (what changes in your business)

Site visits, qualified leads, demo requests, speaking invites, partnerships. These are the metrics that truly matter because they measure real impact. They're slow and hard to attribute.

Vanity metrics live in Output. The mistake is stopping there without ever measuring Outcome.

2. The 8 metrics I watch

OUTPUT (6 metrics)

1. Engagement rate per post (not total)

(Reactions + Comments + Shares) / Impressions, per single post, not aggregated. Looking at monthly average engagement hides differences. 10% of your posts perform 5× the rest: you need to understand why, not average.

Target: >5% for most posts, >10% for top performers.

2. Average dwell time

Average time a user spends reading your post. LinkedIn doesn't show it directly but you deduce it from engagement rate on long post vs short post. If a long post has similar engagement to a short one, dwell time is low.

What it tells: if content holds attention after the hook.

3. Comments per impression (not absolute comments)

Comments / Impressions × 100. Because commenting requires much more effort than liking, it's a stronger indicator of perceived value. 30 comments on 10,000 impressions (0.3%) is excellent. 30 on 50,000 (0.06%) is mediocre.

4. Qualitative reactions vs likes

LinkedIn has 6 reactions: Like, Celebrate, Support, Love, Insightful, Funny. "Insightful" is the strongest signal: indicates the reader found value. Percentage of Insightful/total reactions is a proxy of content quality.

Target: >15% of total reactions should be "Insightful" on analytical value posts.

5. External shares vs internal

A publicly shared post (share on own feed) is worth much more than a like. A privately shared post (DM to others) is worth even more because it indicates "this was passed to someone specific". LinkedIn doesn't distinguish the two, but total share volume is a useful proxy.

6. Follower growth per post type

How many new followers a post generates on average by format (long text, carousel, video, etc). This is the metric telling you which format converts best for your specific case.

OUTCOME (2 metrics)

7. Post-content profile clicks

How many people click on your profile after seeing a post. Useful proxy of "generated interest". LinkedIn shows it as "profile views" but it should be correlated to number of posts published in the period.

8. Value conversations activated (manual)

Number of DMs/emails/calls received each week generated by LinkedIn content. This isn't measured automatically: you track it in a spreadsheet. But it's the metric connecting LinkedIn to your real business.

3. The 4 metrics I ignore (and why)

Total impressions

Without context (who did you reach?) it's just a number. 100,000 impressions on unqualified users is worth less than 5,000 on decision makers in your ICP. LinkedIn doesn't give qualitative breakdown, so total impressions is a misleading number.

Total likes

Like is the lowest-cost reaction: a thumb means "I saw". It's not a strong signal of perceived value. Better to monitor qualitative reactions (Insightful, Support) than generic likes.

Total follower count

Vanity number par excellence. An account with 50,000 followers of which 40,000 are inactive is worth less than one with 5,000 followers of which 3,000 are in your target. Looking only at follower count hides base quality.

Generic demographic breakdown

LinkedIn shows breakdown by industry, seniority, geography of your followers. Interesting but not actionable weekly. Useful once a quarter to verify audience composition isn't drifting from your target.

4. Review cadence: weekly, monthly, quarterly

Weekly review (15 minutes, Friday)

Check metrics of the week's 5 posts. Identify best and worst. Note what differentiated them (format, time, hook, topic).

Monthly review (60 minutes, last Friday of month)

Check monthly aggregate. Which format performed best? Which topic pillar generated more engagement? Are there time patterns? Adapt next month's editorial calendar.

Quarterly review (2 hours)

Change strategy only quarterly, not weekly. Weekly fluctuations are statistical noise. Real patterns emerge on 12+ posts. If after a quarter a pillar doesn't perform, change. Not earlier.

5. How to read metrics with AI Agent

LinkedIn's dashboard requires opening 4-5 different views, making manual comparisons, keeping context in mind. An AI Agent for analytics accelerates the process: ask natural-language questions and receive answers with data and recommendations.

Examples of useful questions:

• "Which format performed best this month?"
• "Why did the post from the 15th get triple engagement of others?"
• "Do posts on Monday perform better than on Wednesday?"
• "Which of my 4 pillars is generating most new followers?"
• "Is there a pattern among the 5 posts with most valuable comments?"

An AI Agent responds in 2-3 seconds with aggregated data and context. Without AI, the same questions require 30-60 minutes of dashboard manipulation.

6. When to change strategy vs wait

The most common mistake is changing strategy too early. 2 low-performance posts don't mean your strategy is wrong: they might be out of phase with news cycle, published at sub-optimal time, on topics your audience wasn't ready to discuss that week.

Rule of thumb: 15-20 posts are minimum for drawing conclusions. Below this threshold, don't change.

Signals requiring strategy change (after 15+ posts):

• Average engagement rate < 2% on long text posts (most-rewarded format)
• Negative follower growth (losing more followers than gaining)
• No value conversations generated in 4 consecutive weeks
• Your quarter's top performer is always in the bottom 30% of competitors on your topic

If you see 2-3 of these signals, change. Otherwise keep the strategy, iterate on details (hook, times, format) and wait another 15 posts.