An Argument Against Hashtags

Some industries should consider leaving #hashtags behind as they enter 2020. Is yours one?

Predictive analytics and decision tree modeling may be one of the best ways to audit your current social media strategy. This higher ed case study shows exactly why.

Results summary report (explained below)

Case Study: Instagram Competitor Analysis

In 2019, with a team of student workers at Mount Holyoke College, we gathered qualitative and quantitative data for individual posts from both our own Instagram and 8 competing colleges’ primary Instagram channels. We gathered data for approximately 400 individual posts.

Mount Holyoke's Instagram Channel

We reported on several metrics including:

• Post Date

• Target Audience: alums, current students, prospective students, undefined, etc

• Post Type: student highlight, campus beauty, etc

• Post Qualities: long text, quote, teaser for long form content, etc

• Image Quality

• Number of Hashtags

• Likes

• Comments

• Engagement Ratio: compared to current follower count

Data from all colleges were analyzed using decision tree modeling to figure out what variable had the largest impact on the engagement ratio.

Decision tree report. It was run using both a primary sample and a test sample that was segmented out and tested separately. While it's not pretty to look at, the insights it grants are incredibly insightful.

Results

The largest predictor of engagement was the number of hashtags used. Engagement plummeted when more than one hashtag was used with the exception of campus beauty posts.

Posts using less than 1.5 hashtags had an average engagement ratio of 7.3% while those using more than 1.5 hashtags only had an engagement ratio of 4.5%, nearly 3% less.

Posts with a clearly defined target audience performed best when targeted with video content (engagement ratio of 11.9% vs 5.7% for all other post types).

The Theory

While using hashtags expanded reach to broader audiences, these audiences likely did not find the content relevant or important. When external audiences failed to engage during the first moments of the post’s ‘life,’ it signaled to Instagram that the content wasn’t meaningful and suppressed its reach to subscribed, internal audiences who already like the account, thus hurting overall engagement. While, on face, we’re taught best practice is to use hashtags, the data in this case study doesn’t support theory in American higher ed.

Now What?

This suggests a need to change strategy. These changes can include:

• Use less hashtags

• Better tailor hashtags by using explicit college signifiers (e.g. #wearemountholyoke)

• Use more content specific hashtags, e.g. #RedStorm, #BigEast, or #D1 on St. John’s sports posts

• Better tailor messaging to external audiences

• Be strategic about hashtag placement, reserving them for posts such as campus beauty

• Edit captions to add hashtags after the post has already reached internal, subscribed audiences

Decision tree models can be amazing for predictive statistics. Predictive stats guide strategy and help delegate resources. Why put effort into things proven less than effective? Why keep shooting ourselves in the foot?

When's the last time you took inventory of your post performance?