Today I decided to analyze how hashtag selection affects the number of views that a fintech post gets on LinkedIn. Marketing websites frequently tell people to use popular hashtags so more people see their posts. So I decided to conduct an experiment by collecting post data myself. After collecting the data, I used multiple regression in OpenOffice Calc to analyze the results.
Collecting the Hashtag Data
I collected posts from the feeds for 20 different hashtags. The terms were fintech, banks, venture capital, banking, finance, accounting, investing, private equity, payments, bank, personal finance, wealth, wealth management, investment, digital payments, future, future of work, e-commerce, and contactless.
Then I recorded the number of followers for each hashtag. Some of the tags were much more popular than others. The future hashtag has 24.7 million followers. The contactless hashtag only has 445 followers.
I also recorded the number of posts that appeared under each hashtag. The algorithmic feed for each hashtag can display posts that are several days old. So I switched to the recent feed and recorded statistics for the last hour.
I ended up with two independent variables, the follower count for each hashtag and the number of posts. The dependent variable was the number of reactions on each post, including celebrate, love, and other emotions. There weren’t many comments so I didn’t use that metric as a dependent variable in this analysis.
Ideally, the dependent variable would be the number of views each post received. But it’s not possible to see the total view count on other people’s posts. So I had to estimate the popularity of the posts by counting the reaction totals instead. And if I didn’t have access to that metric, I’d count comments over a longer time frame.
Analyzing the Hashtag Data
I used the built-in linear regression formula in OpenOffice Calc to get the results. The formula for estimating reactions was 0.26 times the number of posts, plus 2.8 * 10^-7 times the number of followers.
Then I plugged in some numbers for scenario analysis. If there were 10 posts with a fintech hashtag, the expected result was 2.6 reactions.
But including another data point, 1 million followers for that hashtag, only increased the expected reactions to 2.88.
This indicates that the number of posts that include a specific hashtag is a much more relevant metric for predicting reactions than the number of people who follow the hashtag.
And the R^2 statistic for this analysis was 0.50, so even counting the number of posts wasn’t a great way to predict the expected number of reactions.
One explanation for this result could be that a lot of people who are following various hashtags on LinkedIn are not very active on the platform. So just because a hashtag has 20 million followers, it doesn’t mean that 20 million people are actively following the news for that topic.
But a large number of recent posts under a specific hashtag indicate that people are currently interested in that topic. This might include people who aren’t following that hashtag specifically, but are following the industry as a whole.
A viewer may be following a popular tag like venture capital, which has 19.5 million subscribers. And if they comment on a post about venture capital, that post might also include other less popular tags. There are many fintech tags that don’t have lots of followers themselves, but could still appear in popular posts.
But there wasn’t a huge difference between the number of posts that used very popular hashtags and those that didn’t. Fintech has 385,000 followers, yet there were 10 posts about fintech and 10 reactions to those posts. Investing, a much more popular term with 14.5 million followers, generated 21 posts and 9 reactions during the same period.
This was an experiment with a small sample size, conducted early on Sunday morning. So it’s possible that another experiment could have different results. But LinkedIn is a platform with an international audience and many people were actively posting while I was conducting this experiment, so I wouldn’t be surprised if an experiment on a weekday produced similar data.
Either way, it does appear that checking the current activity level for a specific hashtag is a better way to select hashtags for your posts than just looking at the total number of followers. At least in fintech.