Obama campaign director in 2012, Jim Messina, advised for PP for last general election: The network campaign should focus on last 72 hours and get between 11 and 13 hits per voter. "Network advertising was aimed at seeking an immediate and emotional reaction," says Rafa Rubio, a professor at Complutense University. "It is useful only when accumulation of impacts provokes a reaction that affects will to vote," he adds. The purpose of ads on Facebook was no longer to involve voter or to take him to a website with data on pensions or unemployment, but to influence him with aggressive messages.More information
- How Cambridge Analytica worked
- Europe is grounded in marketing of data
The magic of Facebook is ability to adapt each message to a very segmented audience. For example: Man, Leonese-or in provinces where he dances last seat-, older than 50 years, who sympathizes with leaders of PP or citizens and listen Pimpinela. This is a simple example. Segmentation can create tiny groups, which target no more than 30 or 50 people. "The target was a very segmented audience, with a previous job of big data," says Rubio. Big Data's work would allow us to define wher football fans or campers in that group are more inclined to vote one party or anor. Facebook offers tuned tools to segment audience and adjust message.
The company received four parties before Catalan elections
"Every quarter changes everything and what it served recently, now is no longer worth," says Marc Elena, CEO and co-founder of Adsmurai, a Facebook partner for ad optimization. Adsmurai is one of only two Spanish companies that has access to internal Facebook data to make microsegmented ads according to ir audiences. The company, based in Barcelona, received four political parties before Catalan elections: "They wanted to know how y could do it to send message y wanted on Facebook," he says. They didn't work with everyone, but y did. Your clients don't let you say which ones.
It doesn't give more details, but it's easy to imagine. The user's online behavior assigns him to a category. The party creates specific messages for each category segment. Adsmurai can test several of those ads with a group of users. The best-functioning ad is thrown into whole category. Not only that: some party has printed it and put it in mailboxes of voters of neighborhoods with a similar profile. It's publicity on menu. The risk or benefit is obvious: citizens, for example, can segment women close to PSOE but sensitive to permanent prison reviewable. For m, Rivera's message should not be unity of Spain or single contract, but last great crime.
Spanish politics is moving slowly in that direction. The opportunities have been re for years, but tradition weighs. The budgets of campaigns in networks are still scarce. The PP was spent more than 12 million euros in elections of 2016. The heading "Or propaganda and publicity", which includes digital section, had 442,000 euros. While it is true that 100,000 euros in ads on Facebook is a lot of money, difference with or types of advertising is extraordinary.Are you sure this is so terrible?
Cambridge Analytica worked for Trump's campaign in 2016 microsegmenting his messages. They followed example of Obama in 2012. But company presumed innovative because it contributed something more: psychological profiles, with models that predicted character. However, it is not clear that y would end up using those profiles, nor do experts believe y are particularly effective: personalized marketing According to your psychology has shown little effect. Moreover, persuading voters is more complicated than selling m facial cream.
Persuading voters is more complicated than selling m facial cream
We also know that Cambridge Analytica is not foolproof because he first worked for Ted Cruz, Republican candidate who lost primaries precisely against Trump. The importance of company in Trump's victory is likely to have been exaggerated.
There is no scientific evidence to show that political campaigns in networks are very effective. The main difficulty for microsegmentation to be decisive is fierce competition. Campaign messages fight to begin with campaigns of rival parties, using similar techniques. In 2016, for example, Trump used more social media publicity, but Hilary Clinton better exploited e-mail message, according to Fordham University's Jessica Baldwin-Philippi. In addition, se campaigns compete with or things: with electoral debates, television commercials and against all political information. "The factors that decide outcome of an election are multiple and complex," says Frederike Kaluner, head of Privacy International data program, adding: "Measuring and determining effect of a particular variable is exceptionally difficult."The real danger
Facebook and Google have a lot of information about ir users. One of us has downloaded data that Google has on it and re are disturbing details: The company for example stores coordinates of all its searches in Google Maps from 2007. There are exact data of shops, homes, holidays, means of locomotion or workplaces. Google remembers our habits better than us by heart.
It's not just social networks: it's also a medical history, card payments, calls or a mobile trail. The company specializing in risk Experian has credit data of 900 million people around world. The data is an infinite market, especially in United States, where legislation is more permissive and re are files with personal data. The value is not only data itself, but mostly its combination. There are companies dedicated to garing scattered information about a particular person. These ' data brokers ' have data on studies, work, number of children, religion, political ideas, activities, interests, media use and web browsing.
The real dread comes when all se services connect with each or. Facebook fed its profiles with ' brokers ' data such as Acxiom, Epsilon, Datalogix and BlueKai, and latter two were n absorbed by Oracle, anor industry giant. These and or companies serve data so that Facebook can offer more accurate profiles. Facebook warns discreetly: The ads on Web are also decided with "information that have advertisers and our marketing partners." Now Facebook is rethinking se associations: Last week announced that it would stop adding Acxiom variables, dropping a 33% value of shares of that company, according to Reuters.
In Spain re are two companies that use this type of information for policy: Target Point and dialogue. Both have worked with several parties. "We have Spain divided into 36,000 microzones and I can classify 1,000 where snowboarding is made, play paddle or re are many people with private insurance," says Jose Manuel San Millán, Managing Partner of Target Point. "How do matches come to m? There it does go Facebook, Twitter or Instagram, depends on what y use, "he adds. Target Point uses this information primarily to make zonal campaign maps: where to make mailing, put up posters or informative tents.
It's not complicated to build models to find out how likely you are to be diabetic
Where data does not arrive, statistic appears. The information available is used to infer or data, from signs that seem impossible: time that we charge phone, if you do spelling mistakes, speed with which you write or minutes you speak. And worst: all that can be used to infer intimate things. It is not difficult to build models to find out how likely you are to be diabetic, to vote for a party or to be thinking about being a mor.
A famous study of 2013 estimated that it was enough to look at your likes on Facebook to hit your sex in 93% of cases. The model also predicted if you were gay with a 88% probability and your ideology with 85%. These models are in use. Acxiom has in its profiles a variable with probability that any woman is pregnant, and Oracle has audiences of liberal voters or interested in ecology.
In face of this panorama, Cambridge Analytica seems above all a warning of or dangers: "The more you know about someone, and thinner your profiling and easier it is to point out, more options re are that you are able to persuade, influence or manipulate people," says Frederike Kaluner .
All this information can be used to send us personalized marketing or to adapt webs to our tastes. That's assumption. But it can also be a source of discrimination or affect transcendental decisions: granting a loan, denying insurance, offering a contract of employment. Between opacity and speed, networks surpass our ability to understand.