Last year, I launched a social marketing automation app that’s quietly grown to become what I now believe to be the largest third-party “Twitter advertising” platform – I use that term loosely – on the web. To give you a sense of scale, it’s actively generating millions of impressions and hundreds of thousands of engagement events (follows, retweets) across Twitter for users in 41 different countries. Because impressions drive conversions much like a more traditional ad network, this also means it’s spinning off a large amount of data on social engagement and follower interactions. So far I’ve just been letting the data build up, until this past weekend when I decided to dust off my rusty SQL skills and have a look around.
Here are five new things I learned looking at a random data slice of 100,000 Twitter follows.
First, some stats about our overall sample population of followers:
a. The average follower has 3,011 followers of their own on Twitter, although the median follower count is much lower at 388
b. The average follower has a Klout score of 38.9
c. The top 3 most commonly occurring Twitter bio keywords/strings: 1. social (3,857 unique appearances), 2. university (1,940) and 3. media (1,844)
d. The top 3 most common locations (according to their profile): 1. London (2,742 profiles), 2. Boston (2,239) and 3. New York (2,133) (somewhat surprisingly San Francisco came in sixth)
All time for this study is set to Eastern Standard time. Also, with follows being drawn randomly throughout most of the year, our data set shouldn’t need any seasonality adjustments. Nonetheless, if you see any ways to improve this analysis (or any flaws in my methodology, whatever they may be) feel free to shoot me an email.
Lesson #1: Users with Less than 1,000 Followers are Three Times More Likely to Follow You
Trying to get influencers to engage with you on Twitter? The odds aren’t necessarily in your favor, particularly if you’re in the early stages of brand-building for your own account. All other things equal, Twitter accounts with more than 1,000 followers are 62% less likely to follow you than users with <1,000.
Lesson #2: You’re 14% More Likely to Get Followed on a Weekday vs. a Weekend
Out of every 100,000 followers distributed throughout a week, this data says the average weekday will see 14,819 follows vs. an average of 12,952 on a weekend day. That means 14% more follows happen on a typical weekday when people are presumably spending more time on their computers and mobile devices and less time off-line engaging in leisure activities. Overall, approximately 74% of all follows captured in this data set took place between Monday and Friday. The distribution of follows throughout the week was also fairly consistent: Thursday was the highest days for new follows but still only 5.9% above the weekly average, while Sunday following volume was only 10% below average.
Lesson #3: Twitter Engagement is Continuous, with Follower Spikes Around Human Activity Patterns
As this distribution chart of Twitter follows per second shows, Twitter is truly “always on,” with engagement throughout the day looking more like a jagged, organic stream of peaks and valleys than a clear, clean pattern. Interestingly, with 78% of Twitter’s monthly active users (MAUs) outside the US and 76% of MAUs on mobile, almost 71% of daily follows happens outside of 9am – 5pm EST, proving that Twitter’s immense global scale drowns out the typical US work day in terms of driving social interaction. While Twitter engagement seems to be slightly higher in the afternoon (58% of follows happen between noon and midnight vs. 42% between 12:01am and 11:59am), it’s difficult to derive actionable recommendations for your social strategy from such a noisy distribution. One last observation worth making about this particular chart is the obvious spike that happens between 1:00-1:30pm EST. Could this reflect US workers on the east coast returning from lunch breaks and checking their Twitter feed? Could that be coinciding with commuters in Europe getting home from work and checking Twitter around 7:30pm (or children finish their homework after school, then turning to social media)? Impossible to say just from this data-set, but the sharp short-term uptick in follows is hard to miss.
Lesson: #4: Twitter Users Engage Around Social Media, Games, Business and Personal Fulfillment
What are the topics you can talk (re: tweet) about (and/or reference in your bio) that are most likely to get someone to follow you? Here are the four leaders:
1. Social media itself (“social,” “social media,” “social media marketing”)
2. Sports and games (“game,” “games,” “league,” “sports,” “football,” “soccer”)
3. Business (particularly technology and marketing-related topics and keywords)
4. Personal fulfillment and care (“care,” “health,” “spirituality,” “religion”)
In some ways these findings aren’t terribly surprising. As one might expect, Twitter users have a strong positive bias toward social media usage, as well as marketing and technology. High-volume amplification and social TV conversation around live sporting events like the Super Bowl are also pretty expected social media behavior. However, the fourth highest source of social solidarity on Twitter suggests a large share of its users also see social as an extension of their moral compass. Twitter users are highly engaged around topics of health, personal care and the changing — and often polarizing — healthcare landscape. So for those quick to dismiss “viral media” outlets like Upworthy, take heed, they seem to be giving Twitter audiences what they want. And despite Twitter’s secular, Silicon Valley roots, spirituality and religion play a non-trivial role in how users interact with others on the platform.
Overall, talking about one of these four topics is 93% more likely to attract new followers to your account than tweeting about something else, nearly doubling your engagement prospects, according to this analysis.
Lesson #5: Talking About TV Doesn’t Translate to Engagement
One final finding of note is that while social TV represents a relatively large share of voice on Twitter, it doesn’t contribute much to follows, suggesting social TV audiences trend toward broadcasting, not listening, and the quality of conversation is relatively low by engagement standards. In fact, tweeting two or more consecutive posts about a TV show reduces your profile’s rate of new follower acquisition by 39% over the next hour. This lines up with earlier Twitter analysis by Dan Zarella of Hubspot, which showed tweets that include TV social network startup @GetGlue (now tvtag) capture only 1.24% click-through-rates (CTRs), compared to a platform average CTR of 2.11%.
Surprised? Intrigued? Unimpressed? I’d be happy to hear about it if you’d like to drop me a comment.