How Facebook AI may help to change the way we shop online in the future

  • Facebook shares have not recovered from the loss of more than $100 billion after its last earnings report.
  • Facebook has a path back to a record stock price: It needs to monetize the key artificial intelligence algorithms it has been advancing to classify as many as 52,000 traits of users.
  • The most powerful algorithm is called FBLearner Flow: Facebook could use its massive data on user preferences to anticipate the products that consumers want before consumers even realize it, and could work with retailers on predictive shipping.

Facebook has had one tough quarter: data privacy scandals, shrinking margins and a market capitalization that has been trimmed by more than $100 billion as the internet giant's shares tumbled from a record $217 per share down to the $160 range. Facebook now needs to show investors and analysts that it can jump-start earnings growth to restore faith. There is one clear path for Facebook to release its embedded value and regain its mojo: artificial intelligence and predictive shipping.

The current Facebook business model is based upon renting user intelligence to advertisers.

A collage of profile pictures makes up a wall in the break room at a Facebook data center in North Carolina.
Rainier Ehrhardt | Getty Images
A collage of profile pictures makes up a wall in the break room at a Facebook data center in North Carolina.

Regardless of who is mining the data, Facebook eagerly monetizes advertising insights based upon your intimate details. Writer Cale Guthrie Weissman states that "Zuckerberg keeps insisting Facebook doesn't sell our data. What it does is even worse. Renting aggregated data is even more powerful."

As consumer tech radio personality Kim Komando stated, "If it sometimes seems like Facebook knows you personally, that's because it does. It has algorithms that track what you like, watch and click on, and then it passes that information on to Facebook advertisers."

Facebook began selling ads, morphed into big brother and is now the world's largest wholesale renter of AI augmented data. But what if Facebook sold you things before you even knew you wanted to buy them? That may seem like a dystopian fiction, but it already is part of Facebook's ambitions.

As of 2017, Facebook had more than 150 people working in AI and is currently sitting on yottabytes of information about behavior and preferences. According to analysts Julia Angwin and Surya Mattu, "Facebook has tracked more than 52,000 unique attributes to classify users."

In marketing parlance, these are laser-focused psychographics for each user.

The social media giant's 3 key algorithms

Facebook has three new AI algorithms, which will push further into its ability to target users: DeepText, DeepFace and FBLearner Flow.

DeepText analyzes vast quantities of information, much of it from commercial data brokers. According to skeptical analyst Jeffrey Chester from the Center for Digital Democracy, "Facebook is dishonest, bundling a dozen different data companies to target one individual."

Whether you believe the company is being dishonest, building the equivalent of a dozen data companies to target the individual is significant. DeepText can analyze and decipher your own words to direct you toward products you most likely want to purchase.

DeepFace is used to identify people in photos and suggest that users tag people they know. In reality, DeepFace can recognize any face in any photograph on its own. This facial recognition algorithm is actually 97 percent accurate, incredibly even higher than humans who fall a close second at 96 percent accuracy, and the FBI at 85 percent.

But it is FBLearner Flow that is truly terrifying. It is the one that targets you on the basis of decisions you have yet to make. It can accurately implement predictive analytics, knowing which advertisements to show, to which users, for maximum sales uplift. FBLearner knows what you want before you do. The next logical step is predictive shipping, where advertisers send you what you want before you even order.

'Loyalty prediction' — or mind control?

Facebook was on such a roll monetizing AI for precision advertising, investors could see predictive shipping on the horizon. Then something happened. After hitting $217 per share July 25, the stock tumbled after its last earnings report — with the company saying it would need to spend a lot more on security and that would hit profits — and Facebook's challenging summer began.

But the opportunity still exists long-term. Already, advertisers can adjust to new circumstances on the fly and alter your anticipated choices. For example, analyzing nonlinear relationships can predict if you are considering switching from one brand to another. You could then be preempted with advertisements to change your mind entirely.

Facebook's name for this service is "loyalty prediction." My name for this service is mind control.

This is a natural evolution from product recommendations to predictive shipping. Facebook precognition systems can locate products that appeal to individual consumers before they even perceive their own need, whether or not they are ready to buy. By freeing up consumer time, they create happier and more loyal customers. Facebook, their users and shareholders all win.

To be clear, Facebook does not have to recreate the Amazon shipping infrastructure, but could partner with their current advertisers who are already shipping. They are complementary. Facebook is much better suited to smaller retailers with very specific niche markets and Facebook has superior predictive analytics. They can leverage the key competencies of each other. Their advertisers have the shipping and logistics skills. It is a win-win for the last mile to a customer's front door.

Gambling on a bet against Amazon

Predictive shipping would be a big gamble. Facebook is betting that you won't be upset when your new shampoo arrives at your doorstep, whether or not you actually purchased it. Facebook would be entering a new space in the shopping universe, where AI and algorithms take the decision-making out of a customer's hands. Facebook hopes you'll return it, pay for it or take it as a free gift and return for future business. If the shampoo is offered as a free promotional gift, it could build tremendous goodwill. Shopping would become less a deliberate activity and more an organic component of the Facebook daily life experience. Facebook sales increase and margins improve based upon previous searches, purchases, preferences, wishlists and habits.

Facebook would presumably have some serious competition. Amazon was first to patent a service it called "anticipatory shipping" earlier this decade, a concept that would allow the e-commerce giant to deliver more quickly: Analysis of its massive data store would result in an item being sent to a warehouse or truck closer to the destination of the predicted purchaser, but not actually ship to the consumer before a purchase is made. Investors have been placing a higher stock valuation on Amazon as the first mover, early patent holder, but their patent is based upon getting products to staging points like warehouses or trucks. It has taken the Amazon valuation to four commas, or $1 billion, just like Apple.

Facebook predictive shipping goes further to the front door based upon superior predictive analytics. Facebook has been quiet on the topic but maintains a significant lead in the data analytics required to perform predictive shipping, even if it is clearly behind Amazon in logistics. Marius Schober, founder of AR and advertising tech company Advimu.com, believes that by implementing AI assistants, Facebook will be better able to ultrapersonalize and predict needs and wants. "This ultra-personalization will result in even greater conversion [sales and margin] rates," he wrote in a blog post. "This superior Facebook competency is on a faster track to save consumers the effort of thinking."

If after reading this you want to preemptively opt out of Facebook data broker agreements, good luck.

After reading Jeffrey Chester's article on Facebook's use of data brokers, reporter Julia Angwin attempted to opt out from as many Facebook data brokers as she could. She stated that "of the 212 data brokers that I managed to identify, less than half — 92 — accepted opt outs — 65 of them required me to submit a form of identification such as a driver's license."

In the end Julia said, "She could not remove her data from the majority of providers."

— By Keith Wright, Villanova University instructor of accounting and information systems

Latest Special Reports

  • The Edge explores the limitless potential of innovation.

  • A look at 50 private companies set to reshape the business landscape.

  • Automation, AI and machine learning will radically influence the relationship between employers and their workforce.

Tech