B2B Social Media Benchmarks: A 94x Gap on X, and the Ranges You Can Defend

Published engagement rates for X run from 0.03% to 2.83%, a gap of about 94x on one network in one year. Neither figure is a mistake. They divide by different things, count different actions and sample different accounts. B2B social media benchmarks become usable only when you match those three choices, and on X and YouTube almost nobody has filtered for B2B at all.

The short answer

A defensible B2B range today is 5.1% to 5.3% for LinkedIn company pages by impressions and 0.02% to 0.12% for X brand accounts by followers. For YouTube by views it is 3.1% to 3.7%, measured before August 2026. These sit on different denominators, so never rank the three channels against each other.

The B2B social media benchmarks table, with sample and formula

Every row below comes from the publisher's own report, opened on 8 November 2026. Read the formula column before the figure column. Two rows with the same channel and a different formula are not two estimates of one thing.

The B2B column is the one most roundups skip. Only the Oktopost rows are restricted to B2B company pages. Quid's Tech rows are the nearest proxy, and everything else covers all industries or does not say.

Oktopost's Q1 page defers its formula to its PDF, so the formula shown is the one printed in its April 2026 edition. Quid's Tech values were read from the charts in its report, not its text. Socialinsider's intro cites 1.3M LinkedIn posts while its methodology states 2,450,610; this table uses the methodology figure.

Three formula choices that decide an engagement rate

An engagement rate is a fraction, and each publisher picks the top, the bottom and the way to summarise it. The clearest proof is in a single dataset. Rival IQ's 2023 LinkedIn study reported 0.41% by followers and 4.73% by impressions, from the same 58,000 posts. That is about 11.5x before anything about the posts changed, although the two figures also use a median and an average.

The denominator: followers, impressions, reach or views

Followers is a fixed audience you own. Impressions counts displays, reach counts unique accounts, and views counts plays. LinkedIn treats an impression as content at least 50% on screen for 300 milliseconds or more, or clicked, and calls the figure an estimate. Your denominator can change when a platform's counting rule changes, even if your audience does not.

Because a post reaches only a fraction of followers on most networks, a follower-based rate usually runs far below an impression-based one. Publishing them side by side is the most common error in social benchmark roundups.

The numerator: which actions count as engagement

LinkedIn's own analytics count clicks, reactions, comments and shares. Socialinsider and Ordinal leave clicks out. Ordinal adds saves and sends. Socialinsider's X formula counts only likes and reposts, while SociaVault adds replies.

That is why a rate pulled from native LinkedIn analytics should not be held against Socialinsider's 5.20%. You would be counting one more action than the benchmark does. On link-heavy pages, clicks could account for a meaningful share of the total.

The statistic: median or average

Engagement is heavily skewed, so averages sit above medians. Oktopost's own spread shows the shape: a 5.10% median, 8.61% at the 75th percentile, and 21.63% at the 90th. A few strong pages can lift an average without touching the middle.

I would always benchmark against a median. An average tells you what a mixed pool produced in total. A median tells you what a typical account in that pool achieved, which is the question your manager is actually asking.

LinkedIn engagement rate: where the methods can be matched

LinkedIn is the one channel with a real B2B cut, and two datasets on impressions land almost on top of each other. Oktopost's B2B company pages show a 5.10% median for Q1 2026. Socialinsider's 2,450,610 brand-page posts show a 5.20% average for 2025.

Do not read that closeness as confirmation. Oktopost counts clicks, which pushes it up, but reports a median, which pulls it down. Socialinsider excludes clicks, which pulls it down, but reports an average, which pushes it up. My inference is that the two effects roughly cancel, which makes the match partly luck.

The range is also moving. Socialinsider's quarterly data shows every LinkedIn format lower in Q3 2026 than its 2025 full-year figure: document posts 6.40% against 7.00%, text 3.50% against 4.50%, and link posts 2.50% against 3.25%. A 2025 anchor may already be generous, and a page posting mostly links should expect to land nearer the bottom of any range. How the content itself holds up is a separate question, covered in the analysis of which LinkedIn posts still earn attention.

The Technology cut is thin and size-dependent

Oktopost's only published Technology band in Q1 is 201 to 500 employees, at a 5.30% median from 26 to 67 companies per metric in a single month. Its April edition shows top-decile rates falling with size: 21.6% at 51 to 200 employees and 15.2% at 1,001 to 5,000. Treat 5.30% as a reading for mid-sized software firms, not as a sector number. Organic rates also tell you little about paid performance, which is why the comparison of thought leader ads and sponsored content needs its own measurement.

X engagement for B2B: why the published figures differ 94x

The 94x gap is 2.83% from Buffer divided by 0.03% from Quid. Both describe 2025. Buffer's report works from reach under its general rule and draws on accounts that publish through Buffer. Quid divides by followers and samples 150 companies per industry at random.

Socialinsider gives the matched comparison. It also divides by followers and samples brands, and reports 0.12% for 2025, falling to 0.10% in Q3 2026. Against Quid's 0.03%, that is a 4x spread, and the remaining gap has plausible causes: average versus median, organic only versus organic and paid, and a different mix of account sizes.

Same denominator can still produce a 9x gap

SociaVault also divides by followers, yet reports 1.11% across 40,000 accounts. That is about 9x Socialinsider's figure on the same denominator. Socialinsider counts only brand accounts of 1K to 100K followers, and SociaVault stratifies by tier and niche without a brand filter. Formula explains a lot on X. It does not explain everything, and sampling is the rest of the bill.

Technology brands earn about two thirds of the median

Quid's chart puts Tech at 0.02% against the 0.03% all-industry median, so Tech earns about 67% of it (0.02 / 0.03). Buffer adds a warning that matters for a B2B account: it found Premium engagement rose while regular accounts fell, and regular-account medians hit 0% in the latest months. If your brand account is not on Premium, a Premium-influenced median is the wrong yardstick. Before you defend X with a percentage, compare it to what dark discovery does to B2B attribution, since neither number alone says who bought.

YouTube B2B benchmarks: no B2B cut, and a redefined view

YouTube is the hardest of the three to benchmark for B2B. YouTube's analytics documentation lists likes, comments and shares but no engagement rate, so every published rate is a third party's construction. Buffer declines to compute one and reports median views instead, at 433 per video published through its tool. A separate passage gives about 268 for Shorts in 2025, and the report does not reconcile them.

Brands by followers and channels by views are different populations

Quid measures brands by followers and finds Tech at 0.05% against a 0.21% median, the lowest of its 18 industries. Tech therefore earns about 24% of the all-industry figure (0.05 / 0.21), a far deeper discount than on X. CreatorDB and SociaVault both use likes plus comments over views and land at 3.67% and 3.06%.

Their technology rows look reassuring at 3.34% and 3.94%. Those categories are dominated by consumer tech creators and review channels, so I would not hand either number to a B2B software team as a target. SociaVault's Finance and Business row at 2.64% is the closest thing to a business-audience proxy, and it is a topic label rather than a company filter.

The August 2026 view change breaks every view-based series

From 24 August 2026, YouTube counts a view the moment a video starts to play, across Shorts, long-form and live streams. CreatorDB describes the earlier long-form standard as about 30 seconds of playback. A larger denominator mechanically lowers likes-per-view rates, so its 3.67% is a last reading of the old regime.

CreatorDB said it would re-measure in early October 2026. I could not find a published result on 8 November 2026, so no post-change benchmark is cited here. Annotate 24 August in your reporting and treat the two sides as separate series.

The defensible range for each B2B channel

These are working ranges, and the judgement is mine. Each one uses only figures sharing a denominator, states what would count as strong where a percentile exists, and carries a confidence rating. The same discipline applies to any benchmark set, as the comparison of B2B SaaS conversion benchmarks shows for a different metric.

If a board deck needs one line per channel, use the row that matches your export and print the denominator beside the number. An engagement rate without its denominator is an unlabelled unit, and I would not let one onto a slide. Engagement also sits a long way from revenue, a distance examined in the look at LinkedIn's revenue growth and B2B ad spend.

Where this comparison is weakest

The method here is stricter than most roundups, and it still rests on thin ground in two places. Both affect how far you should trust the ranges above.

Almost every source sells software to the people it measures

Oktopost, Socialinsider, Buffer, Quid, Ordinal, CreatorDB and SociaVault are all vendors, and their samples reflect their customers or databases. Oktopost's pages are teams already paying for B2B social tooling. Buffer states its data is not a full-platform view. None of these figures has been independently audited, and the platforms publish no benchmark of their own. These are first-hand datasets, not restatements, but they are not neutral.

The 94x headline overstates the formula effect

Buffer's report gives its X median as about 2.5% in one passage and 2.83% in another, which puts the gap between about 83x and 94x. Its report describes a reach denominator, while its benchmarks landing page prints a follower-based formula, and it never states X's denominator separately. Quid's 0.03% is rounded to two decimals, so the true ratio could be materially different. Treat 94x as evidence that the figures are incomparable, not as a precise measurement of anything.

The strongest counter-argument to this whole post is that matched ranges are still too crude for one account, and I agree. Your own trailing median, computed one way for 12 months, will tell you more than any table here.

Frequently asked questions

Where to start with your own numbers

Export 90 days of posts for one channel this week and compute one number: the median post-level rate, with the formula written beside it. Use impressions on LinkedIn, followers on X and likes plus comments over views on YouTube. Your LinkedIn export will include clicks, so compare it with the Oktopost row, not the Socialinsider one.

Then replace the industry average in your next report with a line showing your own trailing median next to the matching range above. If your YouTube data spans 24 August 2026, split it at that date before you calculate anything. If your LinkedIn team is also running lead capture, keep that measure separate, as the gap between LinkedIn lead gen forms and real conversion explains.

Related reading