[The Trump campaign worked with Cambridge Analytica to] micro-target US voters with carefully tailored messages about the Republican nominee across digital channels.
Intensive survey research, data modelling and performance-optimising algorithms were used to target 10,000 different ads to different audiences in the months leading up to the election. The ads were viewed billions of times, according to the presentation.
…None of the techniques described in the document are illegal. However, the scandal over Cambridge Analytica’s acquisition of data from more than 50 million Facebook users is lifting the lid on an industry that has learned how to closely track the online footprint and daily lives of US voters.
…The Republican nominee, who had just secured sufficient delegates to become the party’s candidate, still had “no speakable data infrastructure” and “no unifying data, digital and tech strategy”, the document states.
…“There was no database of record. There were many disparate data sources that were not connected, matched or hygiened,” she said of the process of ordering, sorting and cleaning enormous data sets. “There was no data science programme, so they weren’t undertaking any modelling. There was no digital marketing team.”
…The document contains very little information about how the campaign used Facebook data. One page, however, suggests Cambridge Analytica was able to constantly monitor the effectiveness of its messaging on different types of voters, giving the company and the campaign constant feedback about levels of engagement on platforms such as Twitter, Facebook and Snapchat.
The feedback loop meant the algorithms could be constantly updated and improved to deliver thousands of different messages to voters depending on their profile.
…Voters in areas where people were likely to be Trump supporters were shown a triumphant-looking image of the nominee, and help finding their nearest polling station.
Those whose geographical information suggested they were not fervent Trump supporters, such as swing voters, were shown photos of his high-profile supporters, including his daughter Ivanka Trump, a celebrity from the reality TV show Duck Dynasty, and Dana White, the president of the Ultimate Fighting Championship.
One of the most effective ads, according to Kaiser, was a piece of native advertising on the political news website Politico, which was also profiled in the presentation. The interactive graphic, which looked like a piece of journalism and purported to list “10 inconvenient truths about the Clinton Foundation”, appeared for several weeks to people from a list of key swing states when they visited the site. It was produced by the in-house Politico team that creates sponsored content.
…Advertisements on Facebook, Twitter, Google and the music-sharing app Pandora were used to help convince 35,000 supporters to install an app used by the most active supporters.
According to the presentation, Cambridge Analytica and the Trump campaign also used a new advertising technique offered by Twitter, launched at the start of the election year, which enabled clients to kickstart viral tweets.
The “conversational ads” feature was used to encourage Trump’s followers to tweet using a set of pre-determined hashtags.
The campaign also took advantage of an ad opportunity provided by Snapchat, enabling users to swipe up and immediately see a preloaded web page. While not useful for securing donors, Cambridge Analytica deemed the tool useful for engaging potential voter “contacts”, according to the presentation.
One of the final slides explains how the company used paid-for Google ads to implement “persuasion search advertising”, to push pro-Trump and anti-Clinton search results through the company’s main search facility.