All Press Coverage

By | Uncategorized | No Comments



Predicting the U.S Election

Without calling a single person’s phone or using another pollster’s data


Predicting Brexit

Correct up to the day of the vote




MIT Enterprise Forum Award

Awarded in 2015 after successfully predicting the Chicago Mayoral Race



Post-Election Interview

U.S Election 2016

Methodology at a Glance

A one minute overview of our technology

Case Study

A deeper, technical presentation done in collaboration with an Illinois Representative and the University of Chicago’s Master of Science in Analytics Program. Reach out for the full case study.




Super Bowl LI predictions, anyone?

By | Featured | No Comments

Super Bowl LI

Over the past week, there has been an abundance of social media chatter about which team will take the 2017 Super Bowl trophy. This chatter is comprised of everything including predictive analysis by professionals, groups of fans arguing back and forth, and of course the individuals who are simply posting online to show their support (or contempt).

Looking through this all gives us an interesting perspective on which team fans believe are most likely to win. Check out the graph below to see for yourself:

Will this method of congregating and analyzing the predictions of others give us enough insight to know who will take home the trophy? We’ll find out tomorrow!

Who We Think the Next President Will Be

By | Featured | One Comment

Election 2016


We’ll make this short and sweet. Here’s our sentiment graph for the 2016 US Presidential election as of 6pm CST today (11/08/16) before most polls closed. The graph up to November 7th was also given to our friends at WCIA Channel 3 and was reported in the morning news. We’ll also be going on tomorrow at 9:30 am for some post-election talk after all the dust clears and we officially know who our next president is.

A Holistic Approach to the 2016 Presidential Race

By | Technology | No Comments

A holistic approach to this year’s election

Our data scientist, Mariah, has been working on some cool stuff for the presidential election. Specifically to do with correlations and comparisons to real world events and testing our graphs against traditional polls. We’re always open to constructive criticism and it’s how we grow as a company. Here’s what she had to say:

The main goal of my work this summer with Tweetsense is to compare their results to polls for the upcoming election. I run Tweetsense’s algorithm each day and then once polls are released, I record the poll results and can compare the two to see if the two considerably different methods yield similar results. There are a few caveats to account for, however. First, the results from Tweetsense are on a scale of -1 to 1, and strong positive sentiment tweets can cancel out strong negative sentiment tweets, giving an overall average score, so it’s not enough just to look at an average and call it a day (a common statistical proverb)

In polling for favorability, the classical polls give out a percentage of people who view a candidate favorably and a percentage for negative views. Another problem to deal with is that voter representation will never be perfect regardless of the method, lest it be somehow omniscient. As you can imagine, many of Clinton’s supporters are not on Twitter, but a lot of Sanders supporters (who will likely vote for her in November) are on Twitter, and they’re not saying great things about her. If they’re saying anything good, it’s not with enthusiasm. Trump supporters, on the other hand, are more evenly distributed among all ages and many of them are on Twitter. And more on this, third party candidates (Gary Johnson and Jill Stein) don’t have many saying bad things about them because few people know who they are but their supporters are strongly favoring them on Twitter. Tweetsense doesn’t conclude opinions using the volume of tweets, but shows opinion data instead. Therefore, popular opinions show Gary Johnson will sweep everyone, though we know that’s highly unlikely simply because of the way the electoral college works.

Even though there are many practical variables one must consider when looking at Tweetsense’s data (can’t just take any information at face value), it has some wonderful features that polls can’t compete with. For instance, it uses only opinions that are already freely shared online, so nobody is bothered with automated phone calls or online surveys. This also makes it much cheaper and less time consuming because nobody needs to be paid to go out and ask how people feel about different candidates and Tweetsense can run its powerful analysis in literally half of a second. Tweetsense’s method also eliminates polling biases such as question bias or interviewer bias because there are no questions and no people asking them. These people are simply sharing their opinions freely without the intent of telling a polling company how they feel. In addition, the millennials remain active on Twitter and are very vocal about their opinions. As we saw with Brexit, it was mostly the young people who turned it and knowing how they feel is crucial to an election where we typically assume the 40 somethings will determine who is our next president. The young people of the country are increasingly politically aware and voter turnout has never been higher, especially for the 18-29 age group.
Looking at Tweetsense’s data is truly fascinating because depending on the election scandals or events, the candidate’s sentiment changes, sometimes drastically. In the graph below, you can see Tweetsense’s sentiment results for Clinton and Trump since mid-June. I personally find the trends surrounding Bernie Sanders’ endorsement of Clinton to be fascinating. The day or two leading up to that significant event, there was much speculation that it would be happening soon and Clinton’s favorability begins to sharply increase. The day he endorsed Clinton, Trump’s numbers began to fall and Clinton’s kept rising steadily. Something else worth noting is that after the Brexit vote, things started changing dramatically for both candidates. I hypothesize that this is because many of the tweeters who were previously saying negative things about both candidates started tweeting about Brexit and their feelings towards that while the strongest supporters of both candidates perhaps continued voicing their opinion on their preferred candidate, causing the overall trend to be positive for a few days following the historic vote.

Tweetsense’s sentiment predictions as of late. Major world events labelled on the graph for convenience. Work and credit: Mariah Yelenick