How this page was reviewed
Written and reviewed by the Postraid editorial team. We compared the linked source page with the provider information and dated tables shown here. This is editorial guidance, not a hands-on product test unless the page explicitly says otherwise. Last updated .
The short version
A benchmark is a reference point from a particular sample. Match the metric, audience and period before deciding whether your content is ahead or behind.
What a benchmark can and cannot answer
A short-form benchmark summarizes a defined collection of posts or survey responses. It can help a team identify a reasonable question to investigate, such as whether its format mix is too narrow. It cannot tell you what a particular account would achieve by copying the average posting behavior.
Before recording a number, record its definition. You need the platform, format, period, account sample, calculation and aggregation method. A percentage without that context can look comparable to another percentage while measuring something quite different.
Use benchmarks as starting hypotheses. Your own content may serve a specialist audience or a different business objective from the accounts in a broad study. Being below a platform-wide average does not automatically mean the content failed, just as being above it does not establish that the content generated useful customers.
The current cross-platform reference
Socialinsider’s August 2026 study reports average engagement rates of 2.60% for TikTok, 0.45% for Reels and 0.30% for Shorts across 69 million videos from January 2025 through July 2026. Its formulas use followers, but TikTok includes shares while the listed Reels and Shorts formulas use likes and comments.
That methodological difference matters. These are not identical collections of interactions divided by an identical exposure measure. Do not interpret the gap as a controlled experiment showing that moving the same post to TikTok will multiply its business value.
Keep the publication date and data period next to the figures in your notes. A report published in 2026 can include substantial earlier activity. If the publisher revises its methodology or figures, update the comparison rather than silently combining versions in a year-over-year chart.
Engagement needs a denominator
An engagement rate based on followers answers a different question from one based on reach or views. A post shown far beyond its followers can produce a striking follower-based rate. A reach-based rate describes interaction relative to the measured audience reached, subject to the tool’s definition.
Consider a fictional post with 100 counted interactions, 1,000 followers and 5,000 measured views. Dividing by followers gives 10%; dividing by views gives 2%. Neither calculation alone says which interaction types were counted or whether those viewers were suitable customers.
Write the formula in the report instead of relying on a short label. Keep raw interactions and denominators available for checking. If two tools disagree, first inspect their definitions and time windows before assuming one is broken or that the audience changed dramatically.
Duration research suggests questions, not padding
Socialinsider’s TikTok length study lists 15–30-second videos at 6.00% engagement and 120–180-second videos at 11,136 median views. The latter is compared with 1,000 median views for 15–30 seconds, not with the 1–15-second group’s 1,274. The distinction corrects a misleading “eleven times the shortest clips” summary.
The useful lesson is that the metric can change the apparent winning length. It is not evidence that extending a weak explanation to two minutes creates more reach. The posts in each group may differ in topic, account, audience and production quality as well as duration.
Design a length test around a complete message. A concise demonstration and a fuller explanation can be legitimate alternatives if both fulfill their promise. A version padded with repeated text is not a fair test of whether more useful detail helps the audience.
Carousels and videos do different jobs
Buffer’s current format report is titled as a 45-million-plus-post analysis, not the older four-million description. It reports TikTok median engagement of 3.39% for video and 1.92% for carousels or photos. Its Instagram comparison uses engagement per reach, including 6.9% for carousels, so those numbers should not be mixed directly with follower-based rates.
Fanpage Karma’s separate comparison reports TikTok carousel engagement more than 81% above video, with reach only 3% higher. That is a different observed result, not a reason to average the two reports into a universal rule. Their samples and methods need to be considered separately.
For a product team, the first decision is whether the idea needs motion or reader-controlled progression. A mechanism may need a video; a checklist may suit a carousel. Test useful executions of both rather than forcing one idea into an unsuitable format just because a headline declares it the winner.
Frequency is not a guarantee for each post
Buffer’s frequency guide reports up to 17% more views per post for two to five weekly TikToks versus one, with larger figures for higher-frequency groups. Crucially, it also says median views stay fairly flat and the benefit involves more opportunities for breakout posts. That does not support a claim that every post improves with frequency.
Observed frequency groups can differ in resources, experience and content mix. A correlation is not a randomized test proving that a smaller team will gain the same amount simply by increasing output. More production can also reduce review quality if the workflow is not ready.
Choose a sustainable increase, if one is appropriate, and track the full cost of approved content. Compare total useful outcomes as well as per-post metrics. A schedule that produces more views but overwhelms the team or repeats an unclear message may not be an improvement for the business.
Read retention without invented universal thresholds
An opening retention chart can help identify where attention falls away. It should be read with the actual content and the platform’s metric definition. Do not rely on an exact percentage about viewers deciding within three seconds unless the underlying study and calculation are available.
The absence of a verified universal threshold does not mean openings are unimportant. It means your test should be specific. Compare two accurate openings for the same explanation and inspect how the audience responds, rather than declaring every post below a borrowed number defective.
Also consider duration. Finishing a very short loop is not the same behavior as watching a substantial tutorial. Completion, average viewing time and later actions can tell different parts of the story. A single retention score should not replace review of whether the post delivered useful information.
Survey adoption is a different kind of evidence
Wyzowl’s 2026 report describes 266 respondents surveyed in late 2025, split between marketing professionals and consumers. It reports that 69% of surveyed video marketers had made social-media videos. This is survey evidence about reported usage, not an observed census of every business or a controlled return-on-investment experiment.
Keep that distinction when reading broader claims about the popularity of short-form. A respondent saying a format works well is useful opinion data, but it is not the same as measured incremental revenue. A large share of marketers using something does not make it necessary for every product.
Use adoption research to understand the landscape and identify questions for your own planning. Use your operational costs, audience needs and measured outcomes to decide investment. Do not replace those decisions with a popularity statistic stripped of the population that answered the survey.
Watch for changing platform definitions
YouTube changed Shorts view counting from March 31, 2025 to include starts and replays without a minimum watch time, while retaining the earlier measure as engaged views. A chart crossing that change needs to identify which metric it uses. Otherwise, a measurement change can look like a creative improvement.
The general lesson applies across reporting systems: keep a data dictionary and note definition changes. Check whether an export uses lifetime values or a period-specific slice. Confirm whether a field is estimated, rounded or unavailable for some posts before treating missing values as zero.
When comparing destinations, preserve their native metrics as well as any normalized calculation. A shared spreadsheet can make the rows look uniform even when the underlying measurements are not. Clear labels are more valuable than visual consistency that hides important differences.
Build an account-level baseline
Group reasonably comparable posts by purpose and format, then review a consistent observation window. Record raw counts, rates and the number of posts in the group. Use a median to reduce the influence of an extreme outlier, but keep the outlier available for creative review.
For a fictional plant-care app, separate quick problem explanations from detailed setup tutorials. A tutorial may attract fewer casual views while answering a more specific question. Comparing both against one broad engagement target can obscure the different roles they serve.
Avoid dividing the sample into so many tiny categories that each contains only one or two posts. Start with a few useful distinctions and add detail only when the data supports it. A baseline should improve judgment, not create a complicated ranking system with unstable numbers.
Turn one finding into a practical experiment
Suppose the app’s videos repeatedly lose viewers before the actual screen demonstration. The next test could move the visible result earlier while keeping the topic and call to action similar. That is a clearer hypothesis than changing the sound, duration, presenter and posting time together.
Write down what would count as an improvement and when you will evaluate it. Include a comprehension check and relevant audience questions, not only a view total. If the new opening attracts more people but misleads them about the feature, it is not a successful revision.
Repeat cautiously and keep context notes. A promotion, product update or unusual external event can affect results. Treat a small experiment as directional evidence, and resist building a permanent strategy from one exceptional post simply because it fits the benchmark you hoped to confirm.
Connect the learning to production
Maintain a brief with accurate product facts and a record of useful creative lessons. The next batch should reflect those lessons without becoming a collection of near-duplicates. Keep format choice tied to the audience question and review each assembled post before scheduling.
Postraid can support its current product-context reaction, meme and carousel workflow within the available plan limits. This guide does not claim a live trend database, guaranteed benchmark-driven output or built-in signup and sales attribution. Keep separate analytics and attribution methods explicit when using them to assess results.
The goal is not to beat every published average. It is to build a repeatable way to explain the product and learn what helps the intended audience. Benchmarks are useful when they sharpen that process; they become distracting when they turn into unsupported promises about the next post.
A few useful answers.
What is a good engagement rate?
Use a relevant benchmark with the same platform, format, period and formula. Follower-based, reach-based and view-based percentages are not interchangeable. Also consider the post’s purpose and sample size before describing a rate as good or bad.
Should I make every video longer?
No. Length studies compare different observed posts; they do not prove that padding a particular post improves it. Test a complete concise explanation against a genuinely useful fuller version and evaluate the metric that matches your objective.
Why do studies disagree about carousels?
Their account samples, date ranges, metrics and aggregation methods can differ. Read the methods before combining conclusions. Choose a format suited to the idea and use your own comparable tests to decide the mix.
Does posting more improve every post?
The cited frequency research does not establish that. Its discussion distinguishes overall opportunities for breakout results from relatively flat median views. Increase cadence only when you can maintain distinct useful content and adequate review.
Can Postraid automatically turn these averages into sales?
No such guarantee is made here. Use Postraid’s supported creation and publishing controls, then evaluate outcomes through a clearly defined measurement setup. An industry average is not a prediction of your next post’s views or revenue.