How to Write a Successful STEM-related Tweet (Part 4)
The key to stepping up your social media engagement
Introduction
In this fourth and last stop of our Tweet success journey, we’ll tie everything together for the ultimate guide to the perfect Tweet.
Throughout the series, our goal has been to predict the number of Retweets a STEM-related Tweet would receive.
We’ve scrapped data, explored the dataset, and built a machine learning model, and in doing so, we’ve discovered correlations between Tweet features and Tweet success. But we were focusing on our dataset as a whole. Here, we’ll look at the characteristics of only the best!
Data Source
We’ll continue to use the data collected in part 1. To recap, here are the features we have extracted from each data entry:
For user features, we have:
- number of followers
- number of statuses
- number of favourites
- number of times the user is listed
- whether the user is verified
- number of days since the account was created
For Tweet features, we have:
- number of links
- number of mentions
- number of videos
- number of gifs
- number of photos
- number of emojis
- number of words
- number of hashtags
- weekday when posted
- hour when posted
- sentiment
- term frequency–inverse document frequency
I’ll define the best Tweets as those retweeted more times than 80% of all other Tweets in our dataset. We’ll also be focusing on Tweet features, as they are more actionable than user features.
Results
While all of the following features may factor into how successful a Tweet can become, we know from part 3 of this Tweet success series that some are more important than others:
Weekday
Top tweets are usually posted on weekdays, with Tuesday being the most favourable.
Hour
Similar to what we found for weekdays, the hour of post also follows a cyclical trend. Most of the top tweets in our dataset are posted at around 4pm UTC.
Number of Emojis
Top tweets tend to have fewer (usually zero) emojis.
Number of GIFs
Most successful Tweets do not include GIFs.
Number of Hashtags
The most frequent number of hashtags in top Tweets ranges from 1 to 6, with 2 being the most common. With anymore than that, the Tweet may become difficult to read as the hashtags become distracting.
Number of Links
Most popular Tweets have zero or one links.
Number of Mentions
Most top Tweets have no mentions, and the number of top Tweets decreases almost exponentially as the number of mentions increases.
Number of Photos
Most popular Tweets have zero or one photo.
Number of Videos
Most successful Tweets have no videos, but a fair amount have one.
Word Count
The most common number of words in a top Tweet ranges from 25–40.
Conclusion
At the end of our STEM Tweet success journey, we’ve summarized our previous results and dived deeper into the common characteristics of top Tweets!
You can find the full Tweet success series below:
Part 2: Exploratory Data Analysis
Part 3: Building a Tweet Success Predictor
Part 4: Summary of Findings + Further Analysis
References
In addition to the ones linked throughout this article, I wouldn’t have been able to complete this project without the help of these awesome examples and tutorials:
[1] Medium | How to Scrape Tweets With snscrape by Martin Beck
[7] Medium | How to Write a Successful Data Science Article on Medium by Lukas Frei
