Whilst in the past week, I wanted to focus on tips to help you get the most out of Twitter and using Facebook, there is something quite different about today’s post.

How does predicting the future with the likes of Twitter and Facebook sound for you? Personally, I think this is really fascinating.

At over 250 million Tweets sent every day, there is a hell of a lot of information out there. A few smart people have said, that analyzing all these messages can give a significant insight over what will happen in the future.

Let’s take a look at these 5 examples and see, if there is any truth in this.

1.) Twitter Accurately Predicts Politician’s Victory at New Hampshire Primary

This first one is the most recent example. Politician Mitt Romney won the Primary election of New Hampshire last week. The stunning part is that Globalpoint Research, did an in-depth sentiment analysis about Tweets related to the election.

The research firm, in a statement beforehand, predict that this based on Tweets:

Romney would not only have the edge, second place would be a tight call between Ron Paul, Jon Huntsman and Rick Santorum.

The actual results are as follows:

Romney came out on top with 37% of the vote, followed by Paul (23%), Huntsman (17%), Gingrich (10%), Santorum (10%) and Texas Gov. Rick Perry (1%).

Pretty crazy right? Twitter’s prediction beat every expert and national poll by much higher accuracy.  It is especially interesting what this means for the future of elections. Mashable had an interesting comment on the debate, saying that Tweets are much less biased than polling answers. What do you think of this? Is Twitter accurate enough to tell you who the next President is?


2.) Twitter knows how you will be feeling this Friday

Here is another very interesting one for you. Scott Golder (@redlog), from The University of Cornell took a look at close to half a Billion Tweets. What researchers wanted to find out was the following:

“How do our moods and feelings change throughout the day, week and year?”

The interesting part here is that they analyzed messages from 2.4 million people in 84 different countries. This makes a lot of sense, as they could also compare different cultures, with different weekly routines and seasons next to each other.

Here are some of the most interesting findings, some more obvious than others:

  • Daily: The most negativity throughout the week happens on Friday, and (suddenly) disappears in the late afternoon.
  • Weekly: On a daily basis, we are happiest in the morning and then our mood goes downhill. In the late evening it bounces back with another spike in positivity.
  • Seasons: Through comparing different countries, the researches wanted to know if sunlight in different seasons affects our mood. The finding was this: “It’s not how much daylight you’re getting, it’s the relative daylight — whether the days are getting longer or shorter — that makes a difference in positive mood.”

If you ask me, this is pretty exciting stuff! Another insight that is well worth mentioning is that happiness is strongly tied to our weekly structures based on our culture. For example in United Arab Emirates mood patterns are very different as they have a Sunday through Thursday workweek.

And all this, solely from looking at a few Tweets. You can read up on Cornell’s study if you want.


3.) Did Twitter predict the revolution in Egypt?

When in the last year a massive revolution started to rise in Egypt and other Middle Eastern countries, Topsy labs asked themselves:

Could we not have predicted all this based on the hashtags #yemen, #iran and #egypt from Tweets going around?

Whilst Twitter played a major element in the revolution Topsy Labs did a correlation analysis of Tweets mentioned. What’s most interesting is that the graph strongly correlates with actual events marked on the timeline:

I am very curious which other events Twitter will be able to predict based on such analysis. Clearly it is Twitter where news is first broken.


4.) Predict the future yourself with Twitter and Timeu.se

A handy tool, also built by the before mentioned Scott Golder is called TimeUse. And essentially, it allows you to find out what people are doing all day. And when they do it.

I did 2 very interesting searches. For “happy”, the biggest spike happens on Sundays, so apparently that’s when most people are happy.

I then tried “working out”, where the biggest spike is on a Monday. That’s quite interesting as it means most people are back hitting the Gym throughout Monday.

Want to give it a go yourself? Check out Timeu.se and let me know what you can predict about the world.


5.) Hedgefund to make bets based on Tweets – beats the market

As the last example, let’s talk about money for a second. Derwent Capital, founded a new fund, that will make investments solely based on sentiment analysis from Twitter. They put $41 million dollars into the fund and then let the Tweet roulette decide its faith.

What’s the outcome?

“Derwent Capital, the hedge fund that is using Twitter sentiment to make its investments, beat the market–and other hedge funds–in its first full month of trading,” (Lauren Dugan, AllTwitter)

This is pretty powerful if you ask me and goes way beyond anything I could have ever imagined happening through Twitter. Here are more insights about how the algorithm from Derwent Capital works in depth.


6.) Predicting and stopping the spread of diseases with Twitter

I am sure you have come across the Google Flu Trends before. It’s essentially a chart that tells you when most searches are happening for various medical treatments and medicines. It is named to be a good indicator for when and how diseases are spreading.

Some research from Johns Hopkins University recently suggested that Twitter may be even more accurate and faster determining the spread of diseases though. They looked at over 2 billion Tweets. They wanted to find out this:

“Our goal was to find out whether Twitter posts could be a useful source of public health information.”

Some of the findings where very interesting, saying that “a number of users were taking antibiotics to treat the flu, even though antibiotics don’t work on the flu.” The potential of this, essentially meaning that public services could take the right steps to avoid such misunderstandings in the future is huge.

Here are a few very interesting findings of symptoms, treatments and general information coming from these Tweets:

You can also read more about this here.


What else can we make happen with Twitter?

Yes, I have to say, writing this blogpost was one of the most interesting ones ever. Personally, I am convinced there will be lots of other incredible use cases of how Twitter can help us in the future.

It is even something that we consider doing in the future with Buffer. Imagine, there are so many Tweets sitting in the queue, which aren’t yet published. Filtering these and predicting what news feeds will look like is pretty exciting.

Enough of my ideas now though, and over to you. What do you think about these examples of how Twitter predicts what is going to happen? Do you see any other potentials where Tweets could be a good indicator of what will happen in the future?

Looking for a better way to share on social media?

Schedule, publish & analyze your posts across the top social networks, all in one place.

Start a 14-Day Free Trial
Written by Leo Widrich

Co-founder at Buffer.

  • Really interesting points here! I had never really thought about Twitter as a way to monitor public opinion before. If these firms can get accurate and objective research, which must be very difficult, I would be really interested to see what the Twitter-sphere had to say about a lot of things!

    James Mignano

    • Anonymous

      Great point James, I couldn’t agree more! If we can learn about anything just from Twitter analysis, I think it will be massive! 🙂

  • I know that I read something recently about marketers tapping into tweets to read sentiment about particular ads and products. And it’s probably a lot more accurate than what people would report in surveys. 

    But it’s incredibly fascinating – though I suppose not too surprising – that via tweets we’re able to monitor other real world events. I’m anxious to see how companies and organizations analyze and react going forward.

    • Anonymous

      Great contribution here Mandy! Yes, I think marketers will also have a big incentive to look more into this big data trend!

      The things that are able to be derived from it are endless and just like you I am sure excited about what will be next! 🙂

  • Very right!

    We have been working on this for quite some time. We (Colbenson) have found some fabulous examples and situations. 

    Many people spoke on the US Airways that made it to the Hudson river that was pictured by a twitter user, but the interesting thing about that event is that the stock price of US Airways just for that day rose up about 14% to turn back to were it was 48 hours later. Remember that online media and TV took over 20 mins plus to report and the market was open. Here is the data http://www.encontrabilidad.org/blog/wp-content/uploads/2010/10/US-Airways-stock-price-15-01-091.jpg

    Other example we recorded was the announcement of ETA to cease armed actions (the Basque terrorist group in Spain). We detected the event about 30 mins before the announcement. Here is a clip of the activity in the Basque Country that day.

    But these are just short term predictions. We have found out you can go further to predict for example what will be demanded on an online store just by crossing its queries,traffic and products to twitter conversations you can predict what will be most demanded hours and days before. 

    Sorry for the long comment, but I leave so much to say anyhow…;)))

    • Anonymous

      thanks for chipping in Angel, that is very interesting! 🙂

  • Heatherswart

    Fascinating article! I’ve been thinking about this for a while but have never actually gotten round to experimenting!

    I would imagine that the same principles apply to advertising and marketing content ie. Analyzing topical interest and frequently discussed issues would provide marketers with wonderfully useful insights as to how concepts and content can be crafted and honed to be more relevant and dynamic.

  • L RT

    In México the death by plane crash of the Secretary of State was predicted on Twitter a few hours before! 

  • Suhail


  • I know My comment is off topic but I can’t stop me for asking,
    when will The Bufferapp is coming to Google+?

    • Anonymous

      Hi Rahul, great to see you here!

      Yes, Google+ is high up on our list, we are just waiting for the Google+ publishing API to be announced! 🙂 

      • Thanks for such a swift reply, one more question plz
        can we expect a free Google+ intgration or we have to pay for it? 

        • Anonymous

          If buffer follows the strategy they’ve followed up to now, we can expect that the basic functionality will be free-to-use, but there will be a paid, expanded option available for those who are interested in it.

  • Today we have a great example too. Costa Concordia (the Ship that sank in Italy on friday the 13th)  trades as CCL. In theory markets at NYSE were open at 1500 NY time which is 2100 Rome time when first tweets of the accident were recorded.

    This morning at LSE (London Stock Exchange) the company is loosing about 100M$. This will hit NYSE trading too and some Brokers will realize that they could have known on friday if they had count with Twitter data to analyze.This reminds me of the advantage that those having early access to the telegraph had for trading between continents.

  • Predicting the past:

    Leo, it’s only predictive when the results are analyzed ahead of time and an outcome is cited. Your examples illustrate the ability to predict the past; not the future. While interesting, it’s not predictive at all.

  • Anonymous

    Hi Leo, we’ve been looking at polling and twitter predictions. Here’s how Romney looks relative to Santorum. http://tpredict.com/predict.php?predictId=41 And in the French elections how President Sarkozy looks like versus the challenger Francois Hollande. http://tpredict.com/predict.php?predictId=1 Neither are 100% correlated to the opinion polls. But they aren’t far off. Shane

  • Angel Maldonado

    Hi there Leo again. I just thought of this post you wrote now over a year ago. I thought it was interesting to add another incredible example. Not purely on “future prediction” but really hard on early detection.

    As you may have heard there was a fatal train accident last week in Spain (Santiago). Well, we detected the first tweet with an image of the catastrophe at 17:13 PM (GMT) https://twitter.com/just_Ru/status/360115421623504896

    This was 2 minutes before the first news reference from news agency europa press that could not confirm there was victims and had no images. https://twitter.com/europapress_es/status/360116042837655552

    Local hospitals, emergency services and the like would have had such value on that piece of immediate information. Specially now that the media starts to unfold the lack of information at the early stages.

    Nowadays every crisis is “reported” earlier on instant messaging open SM as Twitter. Soon will emergency services feed on this. The sooner the better.

    Kindest regards again