LinkedIn analytics: the KPIs that predict revenue, and the ones that don't
Contents
- What LinkedIn actually shows you, screen by screen
- Impressions or members reached: which number belongs in your report
- The formulas LinkedIn does not calculate for you
- Saves and sends: the two metrics that changed the hierarchy
- Dwell time: the signal you cannot see
- Three metrics I stopped reporting, including one LinkedIn has disowned
- Benchmarks: whose numbers are you comparing yourself to
- Native analytics against paid tools, and exactly where free stops
- From posts to pipeline: the Revenue Attribution Report and the version for everyone else
- Your Monday morning dashboard: thirty minutes, no purchase
- What I don't know
- FAQ
- Sources
LinkedIn analytics are the native measurement screens the platform gives every member: about a dozen numbers per post, grouped under four blocks, Discovery, Profile activity, Social engagement, and Link engagement. Most will not tell you whether your content produced revenue. Two of them, added in September 2025, get closer than anything LinkedIn shipped in the previous decade. I have been on LinkedIn since April 2004, and the reporting surface keeps getting richer and less useful at once: more numbers, weaker ties to a signed contract. What follows: the exact labels, the ratios you must compute yourself, benchmarks with their origin attached, and a dashboard you can build before Monday's first meeting.
In short: LinkedIn's post analytics show Impressions, Members reached, In network, Out of network, Profile viewers from this post, Followers gained from this post, Reactions, Comments, Reposts, Saves, and Sends on LinkedIn. Saves and sends, released September 2025, are the closest proxies for value. Engagement rate is not a revenue predictor: LinkedIn itself now says a high Social Selling Index score "doesn't always correlate with measurable sales outcomes."
What LinkedIn actually shows you, screen by screen
Open any published post and click the impression count at the bottom: LinkedIn's help documentation groups post analytics into four blocks, and the labels below are the platform's own.
Discovery. Impressions, the "number of times your post was shown on LinkedIn." Members reached, the "number of distinct members and Pages that saw your post." Two shares of the total: In network (followers and direct connections) and Out of network (everyone else).
Profile activity. Profile viewers from this post and Followers gained from this post, both formerly limited to company Pages and advertiser reporting.
Social engagement. Reactions, Comments, Reposts, Saves ("the number of times members saved your post"), and Sends on LinkedIn ("the number of times members sent your post to others on LinkedIn").
Link engagement. Visits to links. Video posts add Video views, Watch time, and Average watch time.
Two limits: a rolling 365-day cap on historical reporting, and no bulk export for personal profile analytics, so you copy each post by hand.
Impressions or members reached: which number belongs in your report
They answer different questions, and mixing them is the most common measurement error I see in B2B teams. Impressions count screen appearances, repeats included; members reached counts distinct humans and Pages: 400 people seeing a post, 90 of them twice, is 490 impressions and 400 members reached.
Divide one by the other for frequency: impressions ÷ members reached. A frequency of 1.0 means everyone saw it once. Above 1.3, the post is being re-served, usually from late comment activity. I use 1.3 as a working line, not a published benchmark: LinkedIn reports none, and neither does any study I could verify.
For any report that leaves your desk, use members reached: it maps onto a human who could become a customer; impressions exist only to compute frequency.
The formulas LinkedIn does not calculate for you
The platform shows raw counts; ratios are where the information lives.
| Ratio | Formula | What it tells you |
|---|---|---|
| Engagement rate | (Reactions + Comments + Reposts) ÷ Impressions × 100 | Surface reactivity, weakest of the five |
| Frequency | Impressions ÷ Members reached | Whether the post got a second life |
| Save rate per 1,000 | Saves ÷ Members reached × 1,000 | How many judged it worth keeping |
| Send rate per 1,000 | Sends on LinkedIn ÷ Members reached × 1,000 | How many staked credibility to forward it |
| Profile view rate | Profile viewers from this post ÷ Members reached × 100 | How many wanted to know who wrote it |
One warning on engagement rate: LinkedIn's own definition is broader than the three counts above, including clicks on the post and follows it generates, and vendor studies pick their own denominator, impressions or followers. Pick one formula and never change it mid-year.
Saves and sends are tiny in absolute terms, which is why per-thousand is the right scale: 12 saves on a post reaching 3,000 people is 4 per 1,000, a real number to track; as a percentage it reads 0.4% and looks like rounding error.
Saves and sends: the two metrics that changed the hierarchy
LinkedIn added both to content insights on September 10, 2025, labeled Saves and Sends on LinkedIn: before that date, the two highest-intent reader actions stayed invisible.
A save is a bookmark: someone decided the content has a second use. A send is a private forward through LinkedIn messaging, naming yourself to a specific colleague. Neither costs the reader any social exposure: a reaction is public and cheap, a send is private and expensive.
What they do not measure: identity. You get a count, no name, no company, and LinkedIn has published nothing about how either signal weighs in feed ranking. Blog posts claim saves buy a fixed percentage of extra distribution; every version I traced led back to a vendor blog with no method. I am not repeating it.
Dwell time: the signal you cannot see
Dwell time is the number of seconds a member spends on your post before moving on. LinkedIn does not show it, at any subscription tier.
It is real: LinkedIn documented it in an engineering post dated October 1, 2024, describing an "Auto Normalized Long Dwell Model" that predicts whether a member spends longer on a post than on comparable content. The team's own words on why the old method broke: "Modeling only P(skip) is inadequate because the fixed threshold becomes less effective as user time on the platform increases." LinkedIn reports gains in sessions and time spent, "particularly pronounced among users who tend to passively consume content without explicit interactions," and publishes no percentages.
So you have a confirmed ranking signal you cannot read. The nearest proxies you control: Watch time and Average watch time for video, and for text and documents, high reach paired with low reactions, usually a sign of reading, not skipping.
I will not give you a dwell time benchmark. Numbers circulate, attributed to consultancies that never published their sample, the same reflex I argued against in the case against LinkedIn's AI slop button: a figure without its method attached is not evidence, it is decoration.
Three metrics I stopped reporting, including one LinkedIn has disowned
Follower count. It measures accumulation, not relevance. Ten thousand followers of job seekers and peers can sit next to four hundred who sit on buying committees, and the smaller number is the one attached to a budget. No study backs that, but follower count is the number a competitor reads from the outside, which is why it gets optimized and tells you nothing.
Engagement rate, as a standalone. A ratio of cheap actions over a denominator you do not control, usually benchmarked against the wrong population, covered next.
The Social Selling Index. LinkedIn has effectively withdrawn its own endorsement: on its Sales Solutions page for the index, the company writes:
"A high SSI score doesn't always represent the efficacy of a sales person or correlate with measurable sales outcomes." LinkedIn Sales Solutions, Social Selling Index page, consulted August 2026
LinkedIn goes further, saying the effort to lift the score "can distract people from closing deals and building deep customer relationships." The score is still live at linkedin.com/sales/ssi, updates daily, and sales leaders still use it in quarterly reviews: the vendor that invented it says it does not predict the outcome you care about, and I take that at face value.
None of these three is fraudulent, just measuring the wrong object: a metric that cannot move with your revenue does not belong on a dashboard a chief executive reads.
Benchmarks: whose numbers are you comparing yourself to
The most-quoted engagement figure of 2026 is 5.20%, up 8% year over year, from Socialinsider: 1.3 million posts across 16,645 business Pages, January 2024 to December 2025. That is company Page data, not a personal-profile benchmark.
AuthoredUp's personal-profile study, more than 3 million posts, March 2025 to February 2026, updated June 8, 2026, tells a different story. Median impressions for the top 5% of profiles fell from 13,711 to 6,868, while the all-profile median held near 860. Peak engagement rate is 2.68%, for accounts with 1,001 to 5,000 followers, meaning a director benchmarking against 5.20% concludes they are failing at what is actually top-of-distribution performance.
Format performance splits by population:
| Format | Socialinsider, Pages, engagement rate | AuthoredUp, reach multiplier | AuthoredUp, engagement multiplier |
|---|---|---|---|
| Native document (carousel) | 7.00% | 1.39x | 1.30x |
| Multi-image | 6.45% | not reported separately | not reported separately |
| Video | 6.00% | 0.86x | 0.93x |
| Image | 5.30% | 1.20x | 1.33x |
| Text only | 4.50% | 1.07x | 0.78x |
| Poll | 4.20% | 1.78x | 0.37x |
| Link post | 3.25% | not reported separately | not reported separately |
Polls: best reach, worst engagement, eyeballs and nothing else. Video, promoted as this year's format, sits below 1.0 on both axes for personal profiles, and Socialinsider separately logs a 36% year-over-year drop in video views across all Page sizes.
Outbound benchmark: Belkins studied 14,077 of its own client contacts across 2025 plus 15.1 million via Expandi (June 29, 2026): 26% average connection acceptance, 7.2% reply rate, and 1.3% of connected prospects booking a meeting. A personal note lowered acceptance (25.3% against 27.6%) and raised replies (8.2% against 5.3%): it costs acceptances, buys conversations.
Native analytics against paid tools, and exactly where free stops
Shield Analytics, the default here for seven years, is gone: co-founder Andreas Jonsson posted the wind-down notice on May 19, 2026, quoting "Both Google and LinkedIn have made it clear that we cannot continue operating Shield as it was built." Users got 30 days to export before accounts and stored history disappeared, and the shieldapp.ai pricing page now reads "Shield is winding down." Lesson: your data lives on someone else's server and can vanish on 30 days' notice. Whatever you use, export monthly to a file you own.
| Capability | LinkedIn native, free | AuthoredUp | Taplio | Socialinsider |
|---|---|---|---|---|
| Entry price, August 2026 | $0 | $19.95/month, 1 profile | $39/month Starter, $69 Growth, $199 Pro | $82/month Adapt, $124 Optimize, $199 Predict |
| Per-post metrics: impressions, reach, saves, sends | Yes, all of them | Yes | Yes | Page level |
| History beyond 365 days | No | Yes, from install date | Yes, from install date | Yes, from install date |
| Bulk export of personal post data | No, post by post | Yes | Yes | Yes |
| Cross-post comparison and trends | No | Yes | Yes | Yes |
| Competitor benchmarking | No | Limited | Limited | Yes, its main purpose |
| Content creation and scheduling | Scheduling only | Editor and preview | Full, with generative assistance | No |
| Company Page analytics | Yes | Free and unlimited | Partial | Yes |
Everything here can be done at $0: a spreadsheet, one row per post, eleven columns, filled in weekly by hand, about ten minutes a week. What you lose without a paid tool: automatic history, one-click export, competitor benchmarking. What you do not lose: a single metric the ratios need.
Free stops at scale: past four or more people, manual collection breaks down and twenty dollars a month beats the hour it replaces.
From posts to pipeline: the Revenue Attribution Report and the version for everyone else
The Revenue Attribution Report is LinkedIn's own connection to your accounting. It sits in Business Manager, not Campaign Manager, joining your customer relationship management system to your LinkedIn ad activity.
What it connects: Salesforce, Dynamics 365, or HubSpot CRM. What it reports: revenue won, pipeline amount, leads generated, closed-won opportunities, win rate, average days to close, at member or company level, for account-based marketing, targeting named companies, not a broad audience.
The condition nobody mentions: ad activity only, so organic posts never appear in it.
Related: the Conversions API sends conversion data straight from your systems to LinkedIn, not a browser pixel. LinkedIn cites a Dreamdata case study, "up to a 20% reduction in cost per acquisition," from the platform selling the feature. Treat it as directional.
For everyone else, here is the version that costs nothing and works better than expected. Add one required field to your deal record, "how did this conversation start," filled by the person who took the call, and ask it on every first call: "How did you come across us?" Log the answer verbatim. It is imprecise, but it is the only method that catches what tracking cannot, the send: when someone forwards your post privately and the recipient looks you up weeks later, no attribution system connects those two events. The buyer will tell you, if you ask.
A widely circulated figure, 2.74% for LinkedIn traffic converting to leads against 0.69% for Twitter and 0.77% for Facebook, comes from a HubSpot study of more than 5,000 businesses, last updated August 24, 2023, yet still quoted in 2026 as fresh. Your own conversion rate, computed over the last two quarters, beats a decade-old cross-platform average.
Your Monday morning dashboard: thirty minutes, no purchase
Three passes, three different clocks: nothing here requires a subscription.
Weekly, ten minutes. One spreadsheet row per post published last week: date, format, impressions, members reached, reactions, comments, saves, sends, profile viewers, followers gained. Copy from the post analytics screen, then compute frequency, save rate per 1,000, send rate per 1,000, and profile view rate.
Monthly, ten minutes. Sort the month by send rate per 1,000, descending. Write one sentence on what the top four had in common, then the bottom four: the only exercise that changes next month's writing.
Quarterly, ten minutes. Count how many deal records carry a LinkedIn origin in the field you added, then divide their pipeline value by the hours you spent on the platform. That number, dollars of pipeline per hour, is the one worth defending in a board meeting, one nobody else in your market computes.
A caution: three months of data on a few deals is not statistical evidence; I would not restructure a go-to-market plan on one quarter. Four quarters, a different conversation.
The measurement half of this work sits with the publishing half; the frameworks for that are in the All In methods.
What I don't know
I don't know how LinkedIn weights saves and sends in feed ranking. Nobody outside the company does. The platform confirmed the metrics exist and said nothing about their coefficient.
I don't know how much of your current reach is decided by 360Brew, the 150-billion-parameter ranking model LinkedIn's research team described in a paper filed to arXiv on January 27, 2025. As of the last public statements I could find, LinkedIn has not confirmed that feed ranking runs fully on it, and no rollout timeline has been published. People are writing confident articles about "the 360Brew algorithm" that go well past what is documented.
I don't know whether the 2026 reach decline is a permanent reset or a ranking transition still settling. AuthoredUp measured a roughly 50% median drop for top-decile profiles; that is one study, on one population, over one twelve-month window.
I don't know whether saves and sends will stay visible. LinkedIn removed the Creator Mode toggle in March 2024 and reorganized the analytics surface twice since. Metrics arrive and metrics leave. Which is the strongest argument I have for exporting your own numbers to a file you control, monthly, starting this month. When a label changes on those screens, it gets covered here and in the weekly All In newsletter, with the source attached.
FAQ
What is the difference between impressions and members reached on LinkedIn?
Impressions count every time your post appeared on a screen, repeat views included. Members reached counts distinct members and Pages that saw it. LinkedIn's own definitions: impressions are the "number of times your post was shown on LinkedIn," members reached is the "number of distinct members and Pages that saw your post." Divide impressions by members reached to get frequency; anything meaningfully above 1.0 means the post was re-served.
Are saves and sends more important than likes on LinkedIn?
For commercial purposes, yes. A reaction costs the reader under a second and is public. A save means they expect to need the content again. A send means they forwarded it privately to a named person and attached their own credibility to it. LinkedIn added both metrics on September 10, 2025, under the Social engagement block. What LinkedIn has never published is how much either one weighs in feed ranking.
What is a good LinkedIn engagement rate in 2026?
It depends entirely on which account type you are. Socialinsider reports a 5.20% average across 16,645 business Pages, which is not a valid comparison for a personal profile. For personal profiles, AuthoredUp's analysis of more than 3 million posts puts the peak engagement rate at 2.68%, for accounts with 1,001 to 5,000 followers. Benchmarking a personal profile against Page data is the most common measurement error in B2B teams.
What replaced Shield Analytics for LinkedIn reporting?
Shield wound down after its co-founder announced on May 19, 2026 that Google and LinkedIn had made continued operation impossible as the product was built, with users given 30 days to export. AuthoredUp at $19.95 per month for one profile is the closest direct replacement for personal analytics; Socialinsider from $82 per month covers competitor benchmarking. LinkedIn's own screens carry every metric you need, minus history beyond 365 days and minus bulk export.
Can I measure LinkedIn revenue without running LinkedIn Ads?
Not with the Revenue Attribution Report, which lives in Business Manager and only attributes ad activity. The free alternative is self-reported origin: add a required "how did this conversation start" field to every deal record, ask the question on every first call, and log the answer word for word. It is imprecise, and it is the only method that catches private forwards, which no tracking system can see.
Does the LinkedIn Social Selling Index still matter?
It still exists at linkedin.com/sales/ssi and still updates daily, but LinkedIn has withdrawn its own claim for it. The company states on its Sales Solutions site that "a high SSI score doesn't always represent the efficacy of a sales person or correlate with measurable sales outcomes," and warns that chasing the score can distract sellers from closing. When the vendor says its own metric does not predict the outcome, remove it from the dashboard.
Sources
- LinkedIn Help, "Post analytics for your content": exact metric labels and definitions for Discovery, Profile activity, Social engagement, and Link engagement.
- Social Media Today, September 10, 2025: launch of Saves and Sends on LinkedIn in content insights.
- LinkedIn Engineering, October 1, 2024: Auto Normalized Long Dwell Model, dwell time as a feed ranking signal.
- Socialinsider, LinkedIn benchmarks 2026: 1.3 million posts, 16,645 business Pages, January 2024 to December 2025, published March 16, 2026.
- AuthoredUp, "LinkedIn Trends in 2026": 3 million-plus posts, March 2025 to February 2026, updated June 8, 2026, reach and engagement multipliers by format.
- AuthoredUp, "Shield Analytics Is Winding Down": wind-down notice of May 19, 2026, Andreas Jonsson quote, 30-day export window.
- Shield pricing page: current wind-down notice.
- Belkins, LinkedIn outreach study, June 29, 2026: 14,077 client contacts plus 15.1 million via Expandi, 26% acceptance, 7.2% reply, 1.3% meetings booked.
- LinkedIn Sales Solutions, Social Selling Index: LinkedIn's own statement on the limits of the SSI score.
- LinkedIn Marketing Solutions Help, Revenue Attribution Report in Business Manager: supported CRMs, topline, funnel and conversion metrics, attribution models.
- LinkedIn Developers, Conversions and Lead Sync API: Dreamdata case study, up to 20% cost per acquisition reduction.
- Firooz et al., "360Brew: A Decoder-only Foundation Model for Personalized Ranking and Recommendation," arXiv 2501.16450, January 27, 2025: LinkedIn's foundation ranking model.
- HubSpot, LinkedIn visitor-to-lead conversion study, last updated August 24, 2023: 2.74% against 0.69% and 0.77%, more than 5,000 businesses.
- AuthoredUp pricing: $19.95 per month for one profile, August 2026.
- Taplio pricing: Starter, Growth and Pro plans, August 2026.
- Socialinsider pricing: Adapt, Optimize, Predict and Enterprise plans, August 2026.
All In: measure the two numbers your competitors are not looking at
Sends and saves were invisible until September 2025, and most B2B teams still report reactions. The gap between what LinkedIn shows and what predicts a signed contract is where the work is, and it moves every quarter.
All In is the B2B media that decodes LinkedIn, expert blog, weekly podcast and newsletter for SME leaders and sales directors who want to turn LinkedIn into measurable growth. An original creation by Patrick de Carvalho, on LinkedIn since 2004. Motto: "I Never Lose."
Discover All In and get what matters on LinkedIn, every week.