The Long-Term Effect of Engagement Algorithms on Earnings

In the last decade, we have watched internet/social media companies iterate and improve user engagement first through a deluge of A/B tests and now through machine learning techniques.  Powerful computers are collecting mountains of data about a wide variety of behaviors related to your clicks, scrolls, pinch zooms, locations, etc. to best learn how to engage your attention with these companies’ platforms and ultimately increase the value of the ads sold on these platforms.  Their product is not access to your friends’ photos or to superior search results.   Instead, their product is YOU, as they sell your time/attention and their ability to marginally change your behavior.  This paradigm has always been true for ad-supported media from newspapers to radio to broadcast television.  However, algorithms can now test and iterate to incredibly specific stimuli to engage an individual user optimally, with optimally defined as being best positioned to marginally change your behavior (e.g., buying the extra belt that matches the boots that you were not intending to buy until you were served the ad for Thursday Boot Company based on your previous searches for footwear).  We will leave it to others to make judgements for themselves and their families about the merits and pitfalls of these algorithms molding behavior in search, web browsing, social media, stock market trading, gambling, education, fitness, health, etc.  We can certainly imagine ways these algorithms can be used for good and evil.

The rise of market makers that use computer systems to participate in the spread between buyers and sellers has fundamentally altered stock trading.  This new profit pool means brokers can be paid more by dealers/market makers for order flow than they can earn by charging trading commissions.  Discount brokers responded with $0 commissions to create as much order flow as possible, and they have iterated to other strategies that drive engagement and more trading in all types of securities[1].  With this background, Jason Zweig’s column “I Started Trading Hot Stocks on Robinhood. Then I Couldn’t Stop.” in The Wall Street Journal and his follow up column “When the Stock Market Is Too Much Fun” caught our attention recently.  Zweig lays out the clever ways engagement algorithms and gamification are being used to increase user interest/activity and collect stock trading order flow.  Increased retail stock market participation is a hallmark of almost every stock market cycle, and technology innovations and reductions in transaction costs have played a part in growing participation in the past.

Figure 1. Robinhood Screenshots Show Gamification and Other Engagement Tactics

Source: https://twitter.com/petershk/status/1344286419380916228

This application of engagement algorithms to encourage stock trading got us thinking about implications for other industries.  How can engagement algorithms change other industries now and in the future?  Engagement algorithm prerequisites include the ability to gather data on individual user behavior and the ability to personalize the experience for this user.  The data from smart phone apps are readily attainable as devices track your location, clicks, scrolls, pinches, likes, shares, time spent on particular content, and the similar activity of your self-identified friends.  The machines can be starved of much of these data if a user goes to the trouble to limit or block their collection, but most users do not.  In other fields like brick & mortar retail, the collection of data is not as simple, but retailers are innovating.  Map applications are seeking to pair search queries and location data with payment data from credit card companies.  In the future, facial recognition combined with eye tracking and RFID chips will increasingly present opportunities to marginally change user behavior in a customized way in physical places.  Wearables will also play a role.  Smart watches might record when your pulse quickens while viewing an ad or making a particular purchase.  Smart glasses might track your eye movements to see what catches your attention.  Separately, automobiles and appliances have the potential to collect large amounts of user specific data.  Opportunities abound for feeding and then applying engagement algorithms in fitness, medical care, airline travel, workplaces, payments, education, etc.  We are left with a series of questions for any industry we research, including:

• What is currently possible technologically? What will be possible as data collection improves and expands?
• How might that transform the competitive dynamics in the industry? Will the transformation mostly accrue to entrenched industry leaders, to industry disruptors that best understand how to employ engagement algorithms, or to the technology companies that sell the algorithms?
• Will users and/or voters prevent certain industries or companies from collecting data or employing engagement algorithms?

Satisfactory answers to these questions are essential to predicting earnings for 2030-2050 for a wide variety of companies and industries.  Most of our companies are making efforts to enhance their business with these tools.  For example, even though FEMSA’s Oxxo stores have a loyalty program and Mexico’s most popular debit card, most customer purchases go unrecorded.  Under its new digital strategy initiative, however, Oxxo is combining an expanded loyalty program with a digital wallet to better follow and personalize the customer experience.

In general, our portfolio is less exposed to these changes as one of the beverage industry’s appeals is that it is less susceptible to technological obsolescence, making more distant earnings streams predictable with greater confidence.  But for many companies that we research, we suspect that the impact will be greater disruption for the current leaders.  In all cases, we continue to build our research and refine our thinking about how these engagement algorithms may impact our expectations and risks for long-dated owner earnings and cash flows in our valuation models.

[1] Analogous to the social media example above, in many cases a broker’s client has changed from you (i.e., account holders) to the market makers.