Go on Twitter and search up the #GivingTuesday hashtag and you’ll find a never-ending torrent of tweets by non-profits trying to get the attention of digital passersby.
Given how jam-packed the market is, and how short peoples’ attention spans are on Twitter, if there’s anything a non-profit can do to get even ONE like, or ONE re-tweet, or ONE reply on their #GivingTuesday tweets, shouldn’t they learn what that is?
I wanted to try to turn #GivingTuesday tweets into data to figure out the answer to a simple question: Which #GivingTuesday tweets got liked/re-tweeted, and which didn’t? Using my programming skills, I managed to download information about 604,478 tweets, all mentioning the #GivingTuesday hashtag. To my knowledge, this is the most extensive data ever collected on such tweets, ranging all the way from the year #GivingTuesday started, in 2012, all the way to 2020 (may its name and memory be cursed forever…). All in all, we’re talking about 2,436 days’ worth of tweets. Damn!
For each tweet, the information I got includes the text of the tweet, counts of favourites, re-tweets, and replies, info on whether the twitter user included a photo or video, among other things.
In addition to doing some basic analysis on the tweets, I also wanted to use these tweets in a sentiment analysis. That means that I coded as many words as possible in each tweet according to whether or not the word represented one of a relatively small set of sentiments, such as:
This process of coding words for sentiment is, as you can figure out, an imperfect and flawed one. However, I figured that something good and interesting might come out of it, so I forged ahead anyway. However, let’s start by looking at the basic trends of the tweets I collected, then we’ll move on to the sentiment analysis.
#GivingTuesday Basic Twitter Trend Analytics
First off, let’s have a look at how many tweets I was able to download for each year present in the data.
As you can see, before 2018, the data contains anywhere from about 10k to 48k tweets, but then shoots up precipitously in 2019 and 2020! In our current cursed year, we had a total of 2 #GivingTuesdays, between #GivingTuesdayNow, and regular #GivingTuesday.
Yearly trends of engagement metrics
Next, let’s look at the trend in terms of likes, replies, and retweets. You’ll notice in the graph below, I’ve expressed these metrics in terms of the % of tweets with AT LEAST one. I don’t care about virality in this analysis, as there aren’t that many that go viral.
Right off the bat, you see from this graph that tweets became way more likely to be liked/favourited as the years progressed compared to when #GivingTuesday began in 2012. Replies tended to hover just over the 10% point, whereas re-tweets showed a modest upward trend over the years, and then spiked upward in 2019. It looks like it’s been a great time to participate in #GivingTuesday in these last few years!
Engagement by inclusion of photo or video
Now let’s look at something a little bit more practical: How do your chances at getting engagement change based on the media you include with your tweet?
Here we see a VERY clear result: including a video in your tweet is the way to go. Video doubled your chances of getting a re-tweet, increased your chances of getting a like by about 56 percent, and gave you a SLIGHT advantage when it came to replies. Out of the entire dataset of 604,000+ tweets, only 21,595 included a video! That’s just 3.6% of all #GivingTuesday tweets.
Including a video is something almost anyone should be able to do, and is highly valuable, given what we’re seeing here. However, if we ever get to a point where almost everyone is including a video, I predict there wouldn’t be as much of an advantage anymore.
Engagement by # of included Hashtags Continuing along a theme of practical insights, let’s look at our engagement metrics according to the number of hashtags included in the tweet. Have a look below:
On the horizontal axis, I’ve included a count of the number of hashtags included in each tweet. On the vertical axis is the usual “% tweets with 1+…” levels, and the different colours represent likes, re-tweets, and replies, respectively. As much as I hate to admit it, including more than one hashtag seems to help you achieve more engagement from a likes and retweets perspective, but not so with replies interestingly! Likes appear to reach a peak for tweets with 5 hashtags. Re-tweets reaches that peak for tweets with 3 hashtags.
Just please don’t use too many hashtags, for all our sakes!
Engagement by # of included URLs
This is an interesting pattern. Inclusion of even one URL seems to put you at a disadvantage from a likes and replies perspective, but brings your chances to a peak in terms of getting re-tweets. What is it about including a link that makes people shy away from liking or replying? If anyone has a theory, please do let me know!
Enagement by # of mentions
Mention someone on Twitter, and whatever you’re tweeting about gets brought to their attention. That’s why perhaps the below pattern isn’t much of a surprise:
The simple message here is that mentioning more people in your #GivingTuesday tweet basically helps boost your engagement metrics across the board. This effect was especially pronounced in your chances at getting a reply. Re-tweets and Likes, while boosted impressively, were impacted less.
Sentiment Analysis
My aim in this analysis was to try to correlate the relative presence of a small group of sentiments in each tweet with the extent to which those tweets saw engagement of digital passersby. To do this, I had to score each tweet according to the percentage of all of their content words that matched with a known sentiment word. For example, the word “abandoned”, in the dictionary I was using, was associated with the sentiment “sadness”. Also, the word “gift” was associated with the sentiment “anticipation”. Depending on the particular mix of sentiment percentages, a tweet might have been more or less persuasive compared to others.
Which tweets got liked?
According to my analysis, if a tweet was high on “anticipation” words, that tweet had a 62% chance of being liked.
Compare that against the least likely group to be liked: low on anticipation, low in joy, and high on anger. In this group there was a 46% chance of being liked.
Conclusion
This was a lot of fun. I hope you’ve gotten some helpful information out of this, and that this analysis enables you to have the best Giving Tuesday 2021 possible!
To read the full research report including sample tweets to illustrate the most and least effective approaches, click here
Matthew Dubins is the Chief Donor Scientist at Donor Science Consulting. This isn’t just a job for Matthew. Matthew has a Bachelors in Psychology and Sociology and a Masters in Experimental Psychology. His interest in what motivates people converged with his scientific-thinking, methodology, and understanding of data and led him to work at a non-profit where he started honing his understanding and methods of applying data for the benefit of non-profits. Before launching Donor Science Consulting, Matthew worked for the Canadian Breast Cancer Foundation (now Canadian Cancer Society), KCI and Blakely and Cornerstone, where he expanded his knowledge and range of data tools. Matthew saw a need for supporting small- to mid-sized non-profits in their quest for data and realized it was time to open the doors to Donor Science Consulting.
