The benchmark disclosure gap
Three of the 20 most widely cited influencer marketing engagement rate benchmarks publish an engagement rate and disclose all three of the things you need in order to use it: the denominator, the sample size and the date of the data. The other 17 are missing at least one. Nine do not say when their data was collected, and seven do not say what they divided by, which means the figure cannot be compared with any other figure on the list. This audit reads each benchmark at its own publisher page, records what is stated and what is absent, and shows three cases where the gap changes what the number means.
Method: 20 benchmarks were identified by live web search, then each was read at its own publisher page rather than at a page citing it. Fields were recorded as stated, partial or absent. Retrieved 18 October 2026. Full source list below.
What a benchmark has to say to be usable
An engagement rate is a fraction, and a fraction means nothing until you know what sits under the line. The same interactions divided by followers, by impressions, by reach or by views produce four different numbers from one post. So three facts have to travel with any benchmark for it to be usable at all: what it divided by, how many accounts or posts it measured, and when the data was collected.
Across the 20 most cited engagement benchmarks, those three facts travel together three times. The rest arrive with at least one missing, and the field most often missing is the date: nine of the 20 publish no data period whatsoever, including several that carry a current year in the title. A benchmark without a date cannot be aged, and a benchmark without a denominator cannot be compared. Both are still quoted daily as though they were interchangeable.
Three findings
- The most-cited figure is not on the page it is attributed to
Rival IQ's Social Media Industry Benchmark Report is the most frequently named source for brand-account engagement rates. Its own report page carries no absolute median engagement rate for any platform. The only platform-level figures in its key takeaways are year-over-year changes, with Instagram down 16% and X down 48%. The widely repeated Rival IQ medians, 0.36% for Instagram in the 2025 edition and 0.30% in the 2026 edition, could not be confirmed on any Rival IQ page during this pass, and are recorded here as unverifiable at source rather than reproduced as fact. The per-platform medians exist on that page only inside chart images, which carry no extractable text, so the figure may be published without being readable as text. The company is also cited in recent summaries under two names, Rival IQ and Quid, so one dataset circulates under two publisher identities. What the page does disclose, it discloses well: the denominator is interactions divided by total follower count, the sample is 150 companies per industry drawn from a database of over 200,000, covering more than four million posts and nine billion interactions, and the statistic is a median. What it does not state is when any of it was collected.
- A tier grid sourced to a report that does not contain the metric
Favikon publishes a five-platform grid of benchmarks by creator tier, and is unusually careful about one thing: it states a different denominator for each platform, followers for Instagram, views for TikTok and YouTube, reach for Facebook, and warns readers to compare down a column and never across a row. Its sourcing is a single sentence attributing the figures to 2026 industry benchmark data from Influencer Marketing Hub and HypeAuditor. The Influencer Marketing Hub Benchmark Report, read directly for this audit, contains no engagement rate figures at all. It is an attitudinal survey of marketing professionals, with 600-plus respondents, in which engagement appears only as a key performance indicator that respondents selected and as a fraud signal. There is no rate, no tier table and no denominator in it, because measuring rates is not what it does. The other half of the attribution does publish tier figures, but describes them on its own page as rough benchmarks, with no sample size and no collection date. This is how a benchmark becomes unfalsifiable: the grid looks sourced, and following the citation leads to a report without the metric.
- The headline rates are not the same quantity
The widest disagreement on this list is not about creator behaviour. It is about what goes under the division line. Buffer's study of 27 million Instagram posts from 273,000 accounts reports a median engagement rate of 4.3%, and the definition page that study links to gives its formula as total engagements divided by total impressions. Socialinsider's Instagram benchmark reports 0.48%, calculated as likes plus comments divided by the profile's follower count, over 35 million posts from 447,613 pages collected between January and December 2025. Those two figures are roughly nine times apart and neither is wrong, because dividing by impressions and dividing by followers are different operations on the same interactions. Placing them in one table, as aggregated round-ups routinely do, produces a comparison with no meaning. The denominators actually in use across the 20 are followers, impressions, views, reach and, on YouTube in one grid, subscribers. The numerator varies too: Modash divides by followers but counts median likes only, excluding comments, so its figures sit structurally below every other follower-based benchmark even on a matching denominator.
How this was counted
What counts as disclosed
A field is disclosed when the value is stated on the public page: the denominator named, the sample given as a number, the data period given as a date or range. A formula defined elsewhere counts only when the benchmark page links to it. Partial statements, such as a sample floor of at least 100 accounts, are recorded as partial and excluded from the strict counts.
What absent means
Absent means absent from the public page, which is the page that gets cited. Four of the 20 hold figures behind a paywall or a download form, and that is recorded as gated rather than as non-disclosure. One has its figures masked behind a subscription, on a survey period of 2017 that is still quoted today.
What could not be read
Three publishers put their tier tables in chart images or render them client-side, so those values are not present as text and are recorded as unreadable rather than guessed. Two pages blocked retrieval outright, returning an authentication wall and a server error, and are recorded as blocked rather than quoted.
How the ordering was built
There is no public citation index for marketing statistics, so rank is a proxy, not a measurement. It was built from how often each publisher was named as the source by other pages across the searches run for this audit, plus pickup by aggregators. The top of the list is robust; the middle could reorder.
The best disclosure on the list is not from a household name
One benchmark on this list states, on its public page, every one of the following: the population it measured at 141.7 million creators, the definition of that population, the denominator for each platform including the YouTube exception, the interactions counted and excluded, the exclusion rules, the collection month, and a cell-level reliability threshold under which 275 of 300 benchmark cells qualify as reliable at a minimum of 30 creators each. It also volunteers a caveat against its own numbers. That publisher is IQFluence, which is not an established trade publisher and surfaces mainly through aggregated round-ups rather than trade press. Its figures are not flattering to anyone, which may be the point.
Several of the best-known vendors on this list disclose none of those seven items. That is the finding, and it is uncomfortable in a useful way: disclosure quality here is uncorrelated with publisher prominence. The benchmarks that travel furthest are not the ones that document themselves best. The two most-cited entries are missing a collection date and a sample size respectively, while the most completely documented entry is one most readers will not have heard of.
So the operative test is not who published a benchmark. It is whether the page states what was divided by what, over how many accounts, collected when, and whether the figure is on that page at all or only on a page citing it.
Sources
- Rival IQ, Social Media Industry Benchmark Report (per-follower denominator, 150 companies per industry, four million posts, nine billion interactions, medians; no collection date; per-platform medians in chart images only; the 0.36% and 0.30% medians attributed to it could not be confirmed at source).
- Socialinsider, Instagram benchmarks (0.48%, likes plus comments divided by followers, 35 million posts from 447,613 pages, January to December 2025).
- Socialinsider, TikTok benchmarks (3.85% and 4.20% by views, two million videos from 214,507 profiles, January 2024 to December 2025).
- Buffer, Instagram engagement rate study (median 4.3%, 27 million posts from 273,000 accounts; denominator is total impressions, per the definition page the study links to; no collection date).
- Influencer Marketing Hub, Influencer Marketing Benchmark Report (no engagement rate figures; attitudinal survey, 600-plus respondents, May 2026).
- HypeAuditor, Instagram engagement rate (nano 4-5% described as rough benchmarks; current formula divides likes plus comments by views, with a follower-based fallback; no sample size, no collection date).
- Favikon, engagement rate benchmarks (five-platform tier grid, per-platform denominators stated, no sample size, no collection date; attributed to Influencer Marketing Hub and HypeAuditor).
- IQFluence, Influencer Engagement Rate Benchmarks 2026 (141.7 million creators, September 2026, per-platform denominators, 275 of 300 cells reliable at 30 creators minimum).
- Modash, Instagram engagement rate benchmark (median likes divided by followers with comments excluded, 84 million-plus profiles, trailing two-month window; tier values rendered client-side and not present as text).
- Later, Instagram engagement rate benchmarks (2.88% over 44 million-plus feed posts, no collection date; the tier ladder beneath it carries no source, sample or date and is described as guidelines).
- Hootsuite, average engagement rate (Instagram 4.2%, measured per post against follower count for the industry grid; no sample size, no collection date).
- Sprout Social, social media benchmarks by industry (engagement counts rather than rates, three billion messages across one million profiles, 2024 against 2023; no rate denominator).
- Emplifi, social media benchmarks (TikTok median peaking at 35.9%, Instagram falling from roughly 17% to under 10%; no denominator stated, sample given only as thousands of brand accounts).
- Dash Social, influencer benchmarks (nano 3.8% to mega 1.7%; table and FAQ disagree on the macro figure; no sample size, no collection date).
- Starngage, average influencer engagement rates across platforms (TikTok nano 17.96%, 100,000-plus accounts; two formulas given, neither tied to the published table; no collection date).
- Influencer Marketing Hub and HypeAuditor, Instagram influencer performance and fraud report (1.97% at one million followers, 5.60% at 5,000; no denominator stated; tier table in chart images only).
- Statista, Instagram influencer engagement by follower count (figures masked behind a subscription; survey period 2017).
- Metricool, Instagram Study (24.4 million posts from 375,000-plus accounts; no engagement rate on the public page; full study behind a download form).
- Kolsquare, engagement rate study (approximately 10,000 French influencer accounts, first half of 2018; figures gated, study URL now redirects to a resources index).
- Net Influencer, reporting Hootsuite quarterly benchmarks (total audience denominator left undefined; sample given as a floor of at least 100 accounts).
Figures are reported as published by the sources above, retrieved 18 October 2026. Two attributions in wide circulation, Rival IQ's Instagram medians and one publisher's 2026 tier figures, could not be verified at the publisher's own page during this pass and are stated as unverifiable rather than reproduced as fact. Where a figure exists only inside a chart image or is rendered client-side, that is stated at the point of use. Where a report is gated, absent means absent from the public page. Ordering by citation breadth is a proxy and is described in the method section above. This page links to no competitor and reproduces no third-party report; it summarizes published figures with attribution for industry reference.
Go deeper
KALO IQ states the denominator, the method and the date behind every figure it publishes, and draws its benchmarks from its index of 100M+ creators. The benchmark methodology page sets all three out. The platform benchmarks this audit examines are set out in The 2026 Engagement Benchmark Report, and the reason a wrong benchmark hides a bought audience is the theme of the creator verification gap. The human review behind every profile in the index is described in how KALO IQ verifies creators.
The rest of the series: The Verified Creator Report 2026, Creator Economy Statistics 2026, Why Small Creators Win, Who the Social Buyer Is in 2026, the social commerce report, DTC creator spend in Q1 2026 and the shift to always-on creator programs.