What Is a Good Engagement Rate? Social Media Benchmarks 2026

Most of the target figures circulating in marketing meetings come from blog articles whose calculation nobody remembers. What gets passed around are double-digit targets from the follower era, when reach was still tied mostly to your own follower base. X shows how far apart published benchmarks sit today: Buffer gives 3.6% median engagement per post, Socialinsider gives 0.12% engagement per follower for the same platform. Same platform, two denominators, two orders of magnitude. This article delivers the orders of magnitude together with the data behind them, so you know which number you are comparing.

What is a good engagement rate? The reliable orders of magnitude

For Instagram, the most broadly documented data base comes from Socialinsider. The average reach rate stands at 3.50%, a drop of 12% against the previous year, measured across posts from May 2024 to May 2025; the same analysis puts Facebook at 1.65%. On engagement, Socialinsider reports an average of 0.48% for the full year 2025, 24% less than the year before, calculated across 35 million posts from 447,613 accounts. For the second quarter of 2026 the same analysis gives 0.45%, with declining figures across all three main formats.

Two caveats belong with these numbers. First, they are collected globally. Publicly available DACH benchmarks with a comparable sample size do not exist, so for the German-speaking market you are left with your own data set. Second, studies like these describe correlations, not causes. A format that performs better on average does not automatically make your post more successful.

Why engagement by reach is the more honest metric

Calculating against followers assumes that your followers are the audience for your posts. On no major platform is that still true in 2026. Instagram separates connected reach to your own followers from unconnected reach to people served by recommendation, and the second part is growing. A post that reaches 40,000 people, 38,000 of whom do not follow you, produces a flattering follower-based rate and still says nothing about quality.

Socialinsider itself argues that the calculation based on reach is the more meaningful one, because it measures how relevant a post was to the people who actually saw it. That fits what Instagram head Adam Mosseri said about weighting in January 2025: speaking to Social Media Today, he named watch time, likes and sends as the three most important ranking signals, with sends, meaning forwards by direct message, counting slightly more for content shown to non-followers. How the signals interact in ranking is covered in the article on the Instagram algorithm.

For your reporting that means: sends and saves per person reached belong on page one, likes on page three. Both signals show that somebody found the content relevant for other people or for later. The awkward side effect is that exactly these forwards end up in private channels and never appear as a source in any analytics tool.

Benchmark tables per platform, with the underlying data

The tables below are split by denominator, because reach rates and engagement rates are different calculations. Every figure comes with its measurement period and source. Without those details, a benchmark list is a collection of random numbers.

Reach metrics, denominator followers:

Engagement metrics, denominator differs by study:

Where a source does not state its denominator, its value only holds inside its own data set. The two X rows show what follows from that. Buffer gives 3.6% median engagement for X posts and 6.25% for Threads, Socialinsider gives 0.12% engagement per follower for X. Both figures are correct for their own calculation, and both are worthless against each other. Put them side by side in a report without naming the denominator and you have produced a discussion about nothing.

Even within a single study, caution pays: elsewhere the Socialinsider analysis gives a Facebook reach rate of 1.20% instead of 1.65%, depending on the data cut. When you quote a benchmark, the figure, the date and the data set belong together, otherwise in six months you will be quoting a number the source itself no longer carries.

The LinkedIn gap between personal profile and company page is the only row in the table that leads straight to a decision: anyone planning reach through the company page starts with a handicap no content calendar can offset. The Sprout Social figure is an index median across an international data set, not a DACH value. Comparable surveys for the German-speaking market are not public, so you have to recalculate the gap inside your own account.

Formats vary more than platforms do

The platform average is close to useless for setting targets, because the spread between formats is wider than the spread between platforms. Buffer analysed more than four million posts from its own publishing tool, published in March 2026, based on data from January 2022 to October 2024. The result turns the usual expectation around: carousels lead with 6.9% median engagement, single images come third with 4.4%, Reels come last with 3.3%. On reach the picture flips, Reels reach 1.36 times as many people as carousels and 2.25 times as many as single images.

The table is not a complete ranking: Buffer explicitly lists single images in third place, so at least one further format sits between carousel and single image, one for which no figure has been published.

The opposition between engagement and reach is not a contradiction but a question of goals. Reels bring people into contact with the account, carousels bring the people already there to interact. Measure both formats against the same target and you will shut one of them down without cause. One caveat belongs with the data base: Buffer analyses posts published through its own publishing tool. That is a large sample, but not a representative one of what sits on the platforms, and the period ends in October 2024.

From that follows a practical rule for setting targets: compare carousel with carousel and Reel with Reel. A Reel that misses your account's carousel benchmark is not a bad Reel, it is a badly compared one. Which format is the better choice when is settled in the comparison of Reels and carousels.

Which metric fits which goal

Benchmarks only become useful once it is clear which metric represents the goal. As a rule of thumb: one lead metric per account, diagnostic metrics per goal. The assignment below is a recommended approach, not a study finding.

The split has an uncomfortable side effect: it makes visible that a large part of the effect never arrives in measurable form. Forwards run through direct messages and groups, the click comes later and direct. Classic last-click attribution credits that revenue to another source. What works in practice for mid-sized companies is asking buyers where they first came across you, plus holdout and geo-lift tests; media mix modelling needs more data volume than most DACH companies can supply. How to set this up cleanly in methodological terms belongs in analytics and measurement practice.

How to build the only benchmark that counts

External benchmarks answer the question of whether your order of magnitude is plausible. Targets you derive from your own history. The thresholds below come from practical experience, not from studies.

• Fix the period: a rolling median over the past 8 to 12 weeks, not the average. One viral outlier shifts the average so far that nobody reaches the target for three months.
• Split by format: a separate series each for Reel, carousel, single image, document post. A blended figure hides exactly the differences you are measuring for.
• Freeze the denominator: one formula, documented, identical across all channels. Change the denominator mid-year and the time series is worthless.
• A band instead of a point target: define a range, as a rule of thumb median plus 20% as good and median minus 20% as a case to review. Point targets create discussions about noise.
• Record the date: every number in the report gets a date and a data source. Platforms change their metric definitions without notice.

The principle matches what has long been standard in search engine optimisation: relative development before absolute value. The parallel is worth following, see SEO KPIs and the logic behind them.

Common mistakes when comparing benchmarks

• Mixing denominators: putting values from two studies with different formulas into one table. This is the most common and most expensive mistake, because it justifies the wrong priorities.
• Reading a benchmark as causality: carousels get more engagement on average, so we only produce carousels from now on. The studies measure association, not effect, and your account is not the average.
• Ignoring account size: small accounts structurally achieve higher rates, because the denominator is small. A comparison with a corporate account without size brackets is a comparison between two different things.
• Numbers without a date: benchmarks from 2023 in a report from 2026. Instagram lost 24% engagement in a single year-on-year comparison, so a figure two years old describes a different platform reality.
• Followers as the target: the follower count moves without anything changing in the business, and it is the metric that is easiest to buy.

What this means for your reporting

Put two columns into every report that most templates do not have: the formula and the date of the comparison figure. Everything else follows from that, because with a documented formula the wrong discussion can no longer be had. And plan the target anew as soon as a platform changes its metric definition, which in 2026 happens several times a year. The figures given here are as of August 2026.