You are here
Last January the Bank of England announced the creation of a special team of analysts to study how citizens behave on social media. Its ultimate mission: to predict the economy. In the United States, high-risk funds like Tashtego are beginning to study how investors use Twitter and Facebook in the belief that these data shed light on why a hedge fund invests in particular stocks and bonds. All the traces left on Twitter, Facebook or the applications installed on your computer are stored and analyzed, and your searches and purchases are subjected to the closest scrutiny.
Is it really possible to predict the ups and downs of the economy or obtain more accurate forecasts about consumers' behavior by using this type of non-conventional information?
Digital marketing expert Juan Merodio says that if you want to predict consumers' behavior “it's not simply a matter of gathering their data but knowing how to cross-reference them. Paradoxically, we have vast quantities of stored data but often they are of no use to us. I prefer to talk about small data rather than big data, as it makes more sense to analyze small amounts of data than store vast quantities for no reason".
For Merodio, companies have no option but to start creating teams analysts who can interpret the enormous volume of data circulating on the Internet and create algorithms to identify consumer trends. “You might not be aware of it, but whenever you connect to a public Wi-Fi network, you are transferring everything about yourself, the sites you visit, etc. All of which is very valuable information for companies. You probably also have a lot of applications on your device that allow companies to obtain information about you," explained the expert.
Exmaples of predictive apps and websites
One example is the Google Now application, which offers you information before you have even thought about looking for it, such as news alerts that might interest you or the start of your favorite TV show.
Merodio also mentioned the predictive website of the U.S. company Target, which can reach the conclusion that a woman is pregnant by analyzing what she buys. How does it do this? By comparing her purchases with a list of 25 products that women buy when they are pregnant. It then uses this information to send adverts about specific products to this group.
Predictive websites are likely to take a giant step forward with Amazon's Machine Learning app, which lets developers use historical data to create predictive models. The technology proposed by the company is the same as the one used to predict purchases on its e-commerce website.
The app helps companies to use all the data they have collected to improve the quality of their decisions. It can detect problems with financial transactions, prevent a customer from changing to another company, predict the nationality of guests at hotels to offer them the kind of services they appreciate, etc.
“Amazon Machine Learning is the result of everything we've learned in the process of enabling thousands of Amazon developers to quickly build models, experiment and then scale to power planet-scale predictive applications," said Jeff Bilger, Amazon's Machine Intelligence Manager.
In placing this technology at the disposal of every company, Amazon hopes to position itself in the predictive engine market and compete with Google, Microsoft and IBM. At the Machine Learning presentation, the U.S. giant stressed how easy it is to use. There's no danger in getting lost in a sea of data, even if you have no expertise in statistics or data analysis.
Juan Merodio points out that Spanish companies have still a lot to learn as regards data analysis and stresses the importance of analyzing information "in real time. It's vital for the future of companies. Consumers change very quickly and you have to know how to analyze data on the spot. A company's success is going to depend on predicting data". This dependence suggests that big data experts are going to become the most sought-after professionals of the 21st century.
Más información sobre Big Data y Análisis predictivos