For a full explanation of what The Great Twitter Secret Santa is, and the story behind it, read this post.

A lot of surprising tidbits came out of The Great Twitter Secret Santa project, a number of which I find interesting because I’m fascinated by what can be extracted from a raw set of data.  From the signup trends to the traffic analytics, I learned a great deal about network effects, targeted audiences, and virility of social projects.

Signups

The infographic above (my first ever, so be kind) summarizes our sign-ups and ultimately the matching process.  We had 2,196 people sign up, where sign up is defined as providing an email address and at least starting the Twitter authentication process.  Approximately 645 of all sign ups did not fully finish the sign up process, and thus did not confirm our Terms of Service.  About half of that group did not authenticate with Twitter at all, leaving us with just their email address.

We then identified matches between the 1,550 or so left in our confirmed pool.  A match was defined as a another participant of TGSS that you follow and also follows you.  After determining all possible matches and over 700 did not have matches.  We were a bit disappointed by the large number of people who did not have matches from the onset, but given the nature of relationships on Twitter, it was to be expected.  Prior to this project, I suspected most Twitter users did not connect with the same group of people they would on a network like Facebook, given that it requires both individuals to execute a “linking” action and personalities are more obfuscated with Twitter.  While I can’t say with certainty that this is the case, patterns of matches amongst my group of friends who participated seem to suggest it to be true.

Left with 828 people who had at least one match, we were able to connect 530 individuals with one of their matches.  The remaining pool could not be matched because all of their possible matches were taken by the time the algorithm attempted to choose their secret santa.  I tested several methods for choosing matches, including matching randomly, or with the person with the highest/lowest number of matches, and used the best case scenario, which gave us our final numbers above.

We were hoping to include a higher percentage of those that joined, but a significant portion we had no control over (though we did send multiple reminder emails with match counts) and the remaining unmatched pool was a natural consequence of the selection rules.  As a result, we made the decision to have a “secondary pool”, for which participants were required to opt-in through an email.  The secondary pool ended at 166 people, who we matched according to the number of people they both followed.  It turned out to be a good middle ground to avoid randomness while still allowing a higher number to participate.  Interesting finding related to the secondary pool: far more support messages and deadbeats were observed as compared to the original gift exchange, even though it was roughly a quarter of the size.

Analytics and Coverage

Thanks to the help of a friend, we were the subject of a Mashable article and eventually featured as the headline story of ABCNews.com’s technology section.  As someone who’s spent a fair amount of time on Google Analytics hawking visitor counts and traffic sources for my other sites, I was glued to analytics for hours observing trends and figures.  Below is a screenshot of the traffic numbers we saw over the 10 or so days we ran the project.

Mashable

The Mashable article was our first major piece of coverage, coming about a week after we originally launched the site. While I hadn’t gave it a thought earlier, Mashable is just about the perfect major blog to drive signups; their demographic is comprised of social media directors, general digital enthusiasts, as well as entrepreneurs.  Their Twitter account has over 2.1 million followers – a monumental reach (and on the social network that we needed!).

As you can see above, we converted an astounding 52% of visits from Mashable (those that clicked through the article), as compared to 16% from Twitter and 14% from Facebook.  Having read countless case studies on conversion rates and observing the rates in my own projects, I was shocked at the Mashable numbers, and even fairly surprised at the overall conversion rate of about 20% for the site.

ABC News

The other fairly surprising bit from our traffic numbers came from our coverage on ABCNews.com.  Perhaps we overhyped the ABCNews spot amongst ourselves because it involved video content and ABCNews is such a major household name as we expected rather large numbers on par with Mashable traffic.  Overall, we saw 106 referrals from ABCNews.com to both the Great Twitter Secret Santa and The Great Secret Santa (Facebook version, for charity).  We converted 16%, about on pace with other referring sites, but the overall traffic from the coverage paled in comparison.  I attribute this completely to demographics – the readership of ABCNews technology is a much different crowd than the technology enthusiasts reading Mashable and our traffic logs confirm this.

Other Fun Bits

A little gem we realized about halfway through: in some ways we conducted a social media experiment on a set of people who spend a great deal of time trying to spread their ideas or products through social media.  As someone who has isn’t the biggest fan of the underlying motives behind social media and advertising, it was both interesting and mildly humorous to reverse roles and observe their behavior.  Granted, this is a bit of an assumption based off of a bit of random sampling of the people tweeting about the project, as well as the network of people connected to both Regan and Theo.  But it’s a hunch that has a small amount of backing.

The match determination process examined the followers and those following for each participant and compared those two sets with the participant pool.  A great majority of our signups were from people who had a higher number of followers than the number of people they followed – signaling a strength of presence on Twitter (even after removing outliers like Mashable and other major outlets).  As shown in the infographic at the top of this post, the average participant had 649 followers and only followed 461 (again, Mashable removed from this calculation).  These numbers are somewhat expected as this experiment favored people with strong networks on Twitter, otherwise they wouldn’t be matched given the selection process.

If you made it this far – congratulations on being as big of a data nerd as me – and I hope you found these findings/numbers interesting.  As I said in the last post, this project far exceeded my expectations and I’m really proud of what we were able to throw together.

If there’s anything else you’re curious about regarding the project, drop me a line.  And if you didn’t get matched, had a grinch, or something went wrong, we’ve learned a lot this year and hope to improve by leaps and bounds for next holiday season – so stay tuned!

No tags

<< Don’t Look Up

Debrief: The Great Twitter Secret Santa >>