The Search For Marketing Intelligence In The Age Of AI

Tod Loofbourrow, Chairman and CEO, ViralGains.

What do generative AI, parking in Boston, sliders and marketing data have in common?

The answer lies in a classic Boston dilemma: finding a parking spot.

During my career in artificial intelligence (AI), I’ve had the opportunity to teach AI at several major research universities, including a Harvard course on AI applications which taught the major variations of AI algorithms.

In this class, one of my students became frustrated by the notorious parking situation in Kenmore Square, and for good reason. While Kenmore Square is known for its beautiful walking trails that serve as part of Boston’s Emerald Necklace as well as its numerous shops, hotels and restaurants, this makes it an area where you could often spend inordinate amounts of precious time in your car looking for street parking.

Fittingly, this student decided to create a neural network to predict how long it would take to find a spot. Neural networks process large amounts of data by using interconnected “neurons” that assign different weights to inputs. These weighted combinations are analyzed through scoring and reinforcement functions, which ultimately determine the network’s output.

The student put in lots of data, likely to be predictive of the time it would take to find parking. The model incorporated various predictive factors:

•Time Of Day: Work, school and dinner rush

•Weather: Bostonians hold onto spots longer in snow

•Day Of The Week And Holidays: Different patterns on weekends and holidays

•Time Of Year: School in session versus summer, graduation season, student move-in days

Much to our delight and amusement, the system worked pretty well to predict mean time to park in Kenmore Square…on some days. On other days, it was terrible.

The Missing Factor: The Zero Bit

There was one piece of data—far more predictive than any of the above—that the student had completely overlooked. Every New Englander knows the golden rule of Kenmore Square parking. Say it together with me, baseball fans: “Are the Red Sox playing tonight?”

The Boston Red Sox, the Babe-Ruth-curse-breaking baseball team, plays in Fenway Park, which is located adjacent to Kenmore Square. On game nights, parking is nearly impossible for hours before, during and after the game. Once this simple data point—all 81 home season games and up to 14 additional post-season games (hope springs eternal!)—was added to the model, its accuracy dramatically improved.

While the Red Sox schedule might seem as essential a data input as time of day or day of week, my student—who had recently moved from another country—wasn’t aware of it since the baseball season hadn’t started yet. This underscores how even the simplest factors can be easily overlooked, despite their significant impact on outcomes.

I call this kind of data the “zero bit”—the single most predictive piece of information that transforms an AI model’s effectiveness.

The Lesson For Marketers

This lesson isn’t just for AI models; it’s essential for marketing in the age of data privacy and AI-driven decision-making.

Two key truths stand out:

1. The right data is more valuable than large amounts of data (quality over quantity).

2. As privacy limitations, cookie opt-out changes, regulations and consumer behavior make more data hard to use, each powerfully predictive piece of data becomes far more valuable.

Data may be plentiful, but highly predictive data is increasingly sparse—and it is the core of spending marketing resources on the right prospective customers.

Marketers typically target customers based on:

•Contextual Targeting: What people are reading and where they’re reading it.

•Behavioral Targeting: What they are doing online.

•Historical Data: What they’ve done online in the past (which is becoming harder to use due to privacy restrictions).

However, they may be overlooking the most predictive question—their own version of the zero bit.

Finding Your Zero Bit

Ask yourself—”What question, if answered, would be most predictive of the action we’re looking for?” Then figure out how to find it at scale and how to use it to build your targeting model.

What piece of data best predicts your prospective customers’ likelihood to:

•Consider or purchase your product or service?

•Have a positive perception of your brand?

•Become brand enthusiasts and repeat buyers?

And how can you engage your prospective customers so they want to share this data with you?

With the centrality of AI in our marketing and advertising processes, and the power of zero-party data in transforming how we find and understand customers, the power of the right data—leading to the right audience—is the secret of success.

Find your zero bit, build it into your marketing process and unlock the magic. You might just hit a home run in finding new customers.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?