About this study.
This study is based on 1,100,340 posts across 2,238 accounts from 136 participating organizations on Facebook, Instagram, LinkedIn and Twitter.
The participants represent 24 different verticals across 22 different countries. The time frame of the posts analyzed in study was from February 2017 to March 2019.
It is important to note that the results and benchmarks from this study are observational, and they may not necessarily apply to you or your organization.
For example, we can see a relationship between hashtag use and reach on Facebook where reach goes down as the number of hashtags goes up, but we can’t be 100% sure that the hashtags are the cause. Take these results as a reference and a starting point, but don’t rely on them as a definite prescription for what your organization should be doing on social media. What works for one account may not work for another due to thousands of complex factors, so the best method is to look at your own social analytics, run your own tests and confirm these results for yourself.
About the data export & analysis process
The data in this study was exported by Falcon.io’s Product Manager of Measure. They developed and ran the export scripts to obtain the anonymized post performance data from each social network’s API. The results were written in CSV (Comma Separated Values) files for each social network.
Once post data was exported to CSV files for each social network, the data was imported to visualization software Tableau Desktop by study author Maxwell Gollin. In Tableau, the data was organized to make analysis easier and metrics like engagement rate and total engagements were calculated.
The variables and dimensions in the CSV files were used to create data visualizations with Tableau. Those visualizations were then interpreted and redesigned for this landing page by Jess Lykkegaard, Senior Graphic Designer at Falcon.io. For this study, we did not export any identifiable metadata about the posts such as account ID, post text, attachments, or URLs.
About the social media feed algorithms
To understand the relationship between metrics such as impressions, reach, total engagements and engagement rate, it’s important to look at how Facebook, Instagram, LinkedIn and Twitter’s feed algorithms work. Each feed algorithm is different, but here are the basics that apply to all of them.
Every time you post on one of these networks, the post competes with posts from other users to show up in a user’s feed. Whether your post appears in a user’s feed, and whether it appears at the top or bottom, is decided by algorithms that want to serve users content that is relevant and engaging.
If a lot of users who see that post engage with it, it is more likely to show up in other users’ feeds since the algorithms consider it good content. So, content that gets many engagements or a high engagement rate is more likely to get high reach and impressions.
About the metrics we used
For posting times and days, we based our analysis on the local timezone of the headquarters of the organization which made each post. For example, if an organization had its headquarters in New York City, its posts were assumed to be made in US Eastern Time and to be targeted to an audience mostly in the Eastern Time zone. In cases where the organization headquarters was in a different time zone to the main audience of the organization, this may have created some bias in the data.
Reach vs impressions
When available, we preferred reach as a metric to impressions because reach tells you more precisely how many people saw a post. Impressions can be misleading at times. For example, if a post has 1000 impressions, you don’t know if 100 people saw it 10 times each or 1000 saw it only once. However, reach was not accessible from the LinkedIn or Twitter APIs at the time of the data export, so impressions serve as a substitute for reach on those networks in this study.
In general, we used total engagements as an important measure of how engaging a post was to its audience. This was because rather than looking at likes, comments, shares and other engagement metrics separately, total engagements let us aggregate them and get a composite idea of how much the post drove people to interact with it in any way.
Engagement rate is useful as a metric because it tells us how powerfully a post drove people to interact with it. It essentially shows what percentage of users who saw the post engaged with it. A higher engagement rate generally means more compelling, attention grabbing, or interesting content.
Percentage of followers reached is an important metric because it lets us get an idea of how effective a post was at reaching an org’s followers. It’s scaled based on how many followers a page or account has, so we can think of it as a way to measure post reach that’s independent of audience size.
Medians vs means
Finally, why did we use medians instead of means? When working with data that has extreme outliers, medians tend to be more accurate ways of finding the “center” of the data than means are. For example, if you have five posts, with reach of (100, 200, 300, 400, 10000), the mean would be 2200 and the median would be 300. Looking at each number in the dataset, it’s clear that 300 better represents the “center” of the data in this case. The distribution of post metrics works the same way, with a few large pages skewing the means of the data set disproportionately high. To compensate, we took the median rather than the mean of the key metrics in the study.
About calculating the metrics
Facebook: calculating unpaid reach and impressions
Reach on Facebook is the number of unique users who saw a post. In this study, it was restricted to unpaid reach, meaning only unboosted posts were included in the data to avoid bias. Unpaid reach was calculated by adding nonviral unpaid reach (unique users who saw the post because they follow the posting org’s page) to viral unpaid reach (unique users who saw the post because it was shared by another user). Impressions are the total number of times a post appeared in a user’s screen, and unpaid impressions were calculated the same way as unpaid reach.
Facebook: calculating total engagements
For this study, total engagements were the sum of total reactions + comments + shares + link clicks. This accounted for the most meaningful ways users would interact with a post on the platform.
Facebook: calculating engagement rate
Engagement rate was calculated as total engagements divided by total unpaid reach, multiplied by 100% to give a percentage. This way of measuring engagement rate likely inflated Facebook’s rate relative to Twitter or LinkedIn because total engagements were divided by total impressions on Twitter or LinkedIn. Typically posts have more impressions than reach, so calculating engagement rate with reach would show a higher rate of engagement.
Facebook: calculating percentage of followers reached
For Facebook, this study calculated the percent of a page’s followers reached by a post as unpaid reach divided by total page follower count x 100%. This tells us how many people the post reached relative to how many followers the posting page had. It is a way of looking at how effective the post was at reaching users that is not biased by how many followers a page has. This makes it easier to compare reach between pages with low vs. high follower counts.
Instagram: calculating organic reach and impressions
Instagram, unlike Facebook, does not have data on boosted posts alongside organic post data when exporting from its API. So, all Instagram posts used in this study were organic, meaning they were not boosted posts or paid ads. This means all reach and impression data came from an account’s followers seeing the content in their feeds, or from other Instagram users (non-followers) searching for it or stumbling upon it in the Explore tab.
Instagram: calculating total engagements
For Instagram, total engagements were calculated as likes + comments + saves. This covers all meaningful engagement types available through Instagram’s API.
Instagram: calculating engagement rate
Engagement rate was calculated as total engagements divided by organic reach x 100%. This is the same as the calculation used to determine engagement rate for Facebook. Reach was used instead of impressions as it is a more accurate measure of what percentage of users who saw a post interacted with it.
Instagram: calculating percentage of followers reached
This was calculated as organic reach of a post divided by number of account followers x 100%. The metric estimates what proportion of the posting account’s follower base the post was able to reach. The metric is affected by many variables and feed algorithm factors such as posting time, engagement rate, hashtags usage and more.
LinkedIn: issues with impressions and the API
LinkedIn’s API was unreliable at tracking impressions at any point before March 2018. It is unclear why this was the case, but this meant that in many areas that relied on impression data, only the period from March 2018 to March 2019 was accurate. So, many of the analyses for LinkedIn only apply to the last 12 months of data in the study. Also, all impression data on LinkedIn in this study was from unpaid/organic posts.
LinkedIn: calculating total engagements
On LinkedIn, total engagements were calculated as likes + comments + shares + link clicks. This covers all significant ways of engaging with content on LinkedIn that are available through its API.
LinkedIn: calculating engagement rate
Engagement rate was calculated as total engagements divided by impressions x 100%. This number will be deflated compared to engagement rate on Facebook and Instagram, as engagement rate was calculated based on reach rather than impressions for those networks (which tends to be lower).
LinkedIn: calculating impression to follower ratio
Impression/follower ratio was calculated as post impressions divided by the total followers of the poster’s company page x 100%. This is not as accurate as the percentage of followers reached used for Facebook and Instagram. However, it gives an estimate of what proportion of a page’s followers might have seen the post.
Twitter: impressions, engagements and the API
Twitter’s API is somewhat problematic when it comes to calculating key metrics like engagement rate. This is because for any given post, it only tracks the first six weeks of impressions after posting and only includes organic/unpaid impressions. For engagements such as favorites, replies and retweets however, it tracks both paid and unpaid metrics combined for the lifetime of the post. Though most tweets in the study were unpaid, this could still introduce a bias where engagement rate was artificially inflated by paid + unpaid engagements being divided by only six weeks of unpaid impressions.
Twitter: calculating total engagements
Twitter offers very specific data on tweet engagements in its API. In this study, total engagements were calculated as favorites + replies + retweets + link clicks + hashtags clicks + follows from post + profile clicks. Hashtag clicks are clicks on the hashtags in the tweet to see other tweets using the same hashtags. Follows from post are users who clicked to follow the poster’s account from the tweet. Profile clicks are users who tapped the profile picture on the tweet to view the poster’s account.
Twitter: calculating engagement rate
Engagement rate on Twitter was calculated as total engagements divided by impressions x 100%. As mentioned above, this metric is likely inflated by the fact that Twitter tracks lifetime paid & unpaid engagements together but only tracks unpaid impressions for 6 weeks. However, the metric is also deflated relative to Facebook and Instagram since it was calculated with impressions rather than reach.
Twitter: calculating impression to follower ratio
This metric was calculated as post impressions divided by total account followers x 100%. This gives us an estimate of what proportion of followers a given tweet reached. The metric is likely deflated relative to LinkedIn since Twitter only collects impressions for six weeks after a tweet is posted.
Engagement Based on Character Count
On Twitter, posts with 140-160 characters got the most average engagements at 14, followed by posts from 40-60 characters. Posts with 0-20 characters had the lowest average total engagements.
The trend in total engagements on Twitter followed a U-shaped curve much as the trend in impressions by character count did. So, medium to short tweets from 20-60 characters and medium-long tweets from 120-160 characters got the most engagements on average.
This could be simply because posts of those lengths also had the highest average impressions, so they were more likely to get engagements because they were more often seen by users. It could also be a result of users finding either short, snappy copy or a more full sentence the most compelling to interact with. Whatever the reason, social media pros looking to engage Twitter users should try writing tweets from 20-60 characters or 120-160 characters and monitor post performance accordingly.