May 21st 2014

Selfies are pretty cool, everybody’s posting them. Merriam-Webster just added the word Selfie to the dictionary. The most retweeted tweet ever posted was a selfie from the Oscars.

If you’ve heard the latest novelty dance music hit #SELFIE you’ll recall that the song’s protagonist ponders aloud which filter she should use, X-Pro II or Valencia. After hearing it for the millionth time, it struck me that there is a data-backed answer to that question.

I did a little preliminary investigation and found that there has been no large scale analysis of selfie posting yet done. So I gathered a dataset of just over 160,000 Instagram images tagged with #Selfie and did some analysis. I looked at filters, tags and colors …

Oct 17th 2013

More new ReTweet data! In the past few weeks, I’ve looked at quotes and hashtags and images.

This time I decided to look at Tweet-length, in characters and it’s relationship to ReTweets. The data set is up to 1.4 million randomly selected Tweets, from 1.2 million different accounts.

I found that Tweets between 100 and 115 characters were 34% more likely to be ReTweeted than Tweets outside of that range, with a 99.9% confidence interval. A big drop off in ReTweet probability occurs once Tweets get beyond about 120 characters, but up to that point, the longer the Tweet, the better.

Oct 15th 2013

Continuing my new research into ReTweeting behavior, I also looked at non-alphanumerical characters and their relationship to new-school, native, ReTweets.

Using a dataset now more than 1.2 million Tweets strong, I found two particularly impactful characters, which when present in Tweets tend to correlate with those Tweets being more likely to be ReTweeted: quotes and hashtags.

Tweets that contain one or more hashtags were 55% more likely to be ReTweeted than Tweets that did not. Thanks to the large dataset, there is a 99.9% confidence interval that Tweets with hashtags get more ReTweets than those without.

I also found that quotes had a positive effect. Tweets including quotation marks were 30% more likely to be ReTweeted than those that

Oct 7th 2013

I collected a dataset of more than 400,000 randomly selected Tweets and the number of times each tweet received a “new school” (native) ReTweet. I then compared 4 of the most popular ways to send images to Twitter: Facebook image links (images hosted on Facebook’s CDN, fbcdn-sphotos-h-a.akamaihd.net), Instagram, Twitpic and Twitter’s own, native image uploading service (shown in Tweets as pic.Twitter.com).

Tweets with images uploaded to pic.Twitter.com were nearly twice as likely to be retweeted while the use of Twitpic increased the odds by just over 60%. On the other hand Tweets that used Facebook or Instagram links were less likely to be retweeted. In all four of cases, I found a 99% confidence interval assuring us of the reliability

Aug 14th 2013

One of my favorite topics to research is ReTweeting. I have several very large (millions and hundreds of millions of rows) datasets that I use to do my analysis. The results of my research with these databases are best practices, generated across many industries, timezones, account types and languages. This data is a great starting point for your experimentation.

The best data is always your data. So I created a free tool, ReTweetLab that allows you to do the same kind of analysis that I conduct on my large datasets on your account (or any other Twitter account you want).

For each criteria ReTweetLab analyzes, it shows you two things: what you’re currently doing (the small chart in the text) …